My Tesla Investment Thesis: Why Tesla is the most exciting thing going on in the world today

Introduction


I've been a Tesla investor for nearly 4 years now, but looking back at when I first invested in Tesla stock, I must admit I didn't understand the company all that well at the time. The biggest reason I invested was because of my admiration for Tesla's CEO, Elon Musk. I had just finished reading his biography by Ashley Vance, and I vividly remember finishing the 8th chapter that describes what Elon went through during the global financial crisis of 2008. It's then and there that I decided to invest in Tesla. Reading about what Elon went through and accomplished in 2008, gave me the confidence that even if Tesla isn't guaranteed to succeed, all it needs is a chance at success, because that's all Elon needs. So with that, I bought my first Tesla shares.

From this point onward I started following the stock price, and keeping up-to-date with all the news surrounding the company. So as time passed I learned more and more about Tesla, and I added on to my initial position twice in the following 6 months, because the price dropped a lot, even though in my eyes a lot of good news came out. After these first 6 months of continually following all the news surrounding Tesla, the general automotive industry, and listening to every quarterly earnings call, I had come to understand Tesla a lot better, and I was still very happy with my investment in it. After all, it seemed very obvious to me that all cars would eventually be electric, and that although Tesla was still small, it had a number of massive advantages over the rest of the automotive industry.

During the three and a half years after that I didn't buy anymore stock, although this was due to a lack of funds more than a lack of willingness to invest, because I felt that Tesla progressed very nicely over this period of time. There were a couple of events that blew my mind such as the Model 3 unveiling, and the fact that over 100.000 people reserved the car without even seeing it. Some of them even camped in front of Tesla stores overnight to be able to be first in line. There were also a couple of things that made me pause and think, like when Elon was very rude against 2 analysts in the Q1 2018 conference call, but overall my confidence in my investment remained similarly strong.

Over the past 6-12 months, however, my attitude towards my investment in Tesla has changed a fair bit. It's gone from very positive to extremely positive. The biggest catalysts were the successful Model 3 production ramp, allowing Tesla to be far less reliant on the capital markets to fund their growth, and their Autonomy Investor Day. In this blog post, I'm going to explain why I think Tesla will likely be a future tech giant, and why I think they might even come to rival some of the biggest economies on earth in size.

The Future of Transportation


Before we go into detail on Tesla, we need to establish a few things. First and foremost, to appreciate why Tesla is in such a good position, one needs to understand that in the not too distant future all cars will be electric. As a matter of fact, some parts of the world are only a handful of years away from accomplishing this. The most extreme example is Norway, where in March of 2019, 70% of new car sales were electric (full electric + plug-in hybrid), and 58% were fully electric vehicles. This was a one month peak due to an enormous amount of Model 3 deliveries, but even if you look at the entire 1st quarter of 2019, electric vehicle sales amounted to 61% of the market.


Furthermore, an increasingly large number of governments are creating laws that ban the future sale of Internal Combustion Engine (ICE) cars at some point in the future. Some of these governments are Oxford 2020, Costa Rica 2021, Norway 2025, India 2030, Netherlands 2030, Sweden 2030, London 2030, and China 2040.

If numbers alone are not convincing, and/or you'd like to really understand at a fundamental level why all cars will be EVs, I highly recommend reading this amazing blog post on Wait But Why about Tesla and EVs.

The second thing that one must understand to fully appreciate Tesla's value, is the second revolution currently going on in the automotive industry, namely Autonomy. There are dozens of car manufacturers, tech giants, and startups, currently racing to be the first to create a fully autonomous self-driving vehicle. Google/Alphabet has been at it for over a decade, recently spun off its self-driving vehicle program into a seperate entity, called it Waymo, and raised money at a valuation north of 100B$. Mobileye has been a supplier of Advanced Driver-Assistance Systems (ADAS) to car companies for nearly two decades, was acquired by Intel in 2017 for 15B$, is currently working on a self-driving car, and plans to launch a robotaxi ride-hailing service to compete with Uber as early as 2022. Even Apple has been rumoured to be working on an electric, self-driving car since 2014.

This is just the tip of the iceberg. There are enormous amounts of money being spent to try to develop the first fully autonomous vehicle, and it's not hard to see why. Humans collectively travel 10+ trillion miles over roads each year, and although a ride in a self-driving car will likely cost less than the 2-3$ per mile companies like Uber, and Lyft charge for their rides, it's obvious that this is an enormous opportunity worth trillions per year.

Nobody really seems to agree on when the technology will be ready. Alphabet's Waymo has already launched a mini-robotaxi service in Phoenix, Arizona, but they seem far from ready to scale this worldwide. GM's Cruise has said they will launch a robotaxi service as early as 2019 and Elon has said Tesla will have 1 million robotaxis in 2020, but not all industry experts think the technology will be ready nearly as soon as this. However, most of them do think it's just a matter of time before AI will be able to drive a car more safely than a human can. It's just a question of whether this will take a few years, 5 years, 10 years, or more.

Now that we've established that the future of transportation is Autonomous Electric Vehicles (AEVs), we can start to talk about Tesla specifically. In the rest of this post I will discuss:
  1. Tesla's lead in EVs
  2. Tesla Energy
  3. Tesla Investor Loyalty Program
  4. Autonomy
  5. Elon Musk
  6. Risks
  7. Financials
It's turned into quite a massive post, so grab some drinks and snacks, and get ready for a good read.


Tesla's lead in EVs


I think the only way to go about explaining Tesla's lead in EVs is to divide it up into separate sections, because the number of advantages Tesla has is rather large. So with without further ado, these are the reasons why all other car companies should be shitting their pants, and as soon as they clean themselves up, should get off their lazy asses and get serious about developing compelling EVs.

1: Battery Technology

Most car manufacturers don't make a whole lot of their cars themselves. They outsource the majority of their parts, and besides designing the car, all they really make is the internal combustion engine (ICE)*. We've talked about the future being electric, so this expertise will soon be useless. Instead, what will be the most important technology in the future is batteries. A manufacturer's expertise in battery tech affects not only the range of the car, but also the cost, the fuel (electricity) efficiency, the longevity, reliability, performance, charging speed, battery degradation, and the safety.

*Slight oversimplification. Some make the transmission for example, some don't. But they definitely specialize more in design and assembly, than the manufacturing of individual components.

Tesla has been developing this expertise since the company was started back in 2003, and it shows. The second car Tesla ever developed was the Model S back in 2012 with a range of up to 265 miles. As I am writing this in 2019, the non-Tesla EV with the highest range is the Hyundai Kona Electric with 258 miles. To this date, no car manufacturer has made an EV with a longer range than Tesla's Model S from seven years ago. Next year, Porsche and Volvo are both releasing a car with up to 275 miles of range, barely beating the 2012 Model S, but Tesla's Model S now has a range of up to 370 miles, and Elon has said they will soon produce cars with 400+ miles of range.

Tesla destroys the competition equally hard on the other metrics. For the longest time the Model S P100D was the production car with the fastest 0-60 mph acceleration time out of ALL production cars. Currently it is number two with 2.28 seconds, not far behind the Porsche 918 Spyder with 2.2 seconds, which costs 845,000$, and of which only 918 have been made. Maybe the most important metric besides range is efficiency. Tesla is the leader on this front as well, meaning Tesla is able to squeeze more range out of the same size battery as their competitors:



Tesla's lead in battery tech is worrisome for competitors, but what is really frightening for them is the fact that none of them even seem to even care. When you look at all the companies that are producing batteries for EVs, the only other car manufacturer that makes their own batteries is BYD, a Chinese car company. Volkswagen, on May 15th 2019, announced that they will start producing batteries on their own, but they're only investing 1B€*, Tesla has a 16 year head start**, and Volkswagen is still going to be reliant on other suppliers for the vast majority of their EV battery needs.

*Tesla is spending 5B$ on their first of many Gigafactories
**I don't mean to imply head start is the same as lead. It should take Volkswagen a lot less than 16 years to catch up to Tesla in terms of battery tech.

Meanwhile, other car manufacturers seem content to rely on suppliers completely. Looking at EV specs, none of these suppliers seem to be able to produce batteries as good as Tesla's, but even if the difference in specs between Teslas and other EVs is due to other factors, Tesla is about to get a big step up on the competition thanks to their recent acquisition of Maxwell Technologies.

Maxwell has recently made breakthroughs in a cutting-edge new way to produce batteries, and they see the potential for this new technology to improve even further over the coming years. If you're interested in learning out more about the advantages Maxwell will offer Tesla over the coming years, I highly recommend checking out this video, or one of the other numerous YouTube videos on the subject.

Even if EV battery suppliers are somehow able to keep up with these new technologies, car manufacturers face the even bigger issue of battery supply.

2: Battery Supply

In February of 2014 Tesla announced their plans for the Gigafactory, a gigantic battery production facility, that will be the world's largest building by footprint once completed. Back in 2014 the world's total lithium-ion battery production capacity was 35 GWh per year, consisting mostly of cell phone and laptop batteries. Tesla realised that to achieve it's goal of producing 500.000 EVs by the end of the decade, it would need approximately 35 GWh worth of batteries, or in other words, all of the world's batteries being produced at the time. So they came up with the radical solution of making the world's biggest battery factory, the Gigafactory 1 in Nevada.

The Gigafactory's planned total output is 105 GWh of battery cells, and 150 GWh of battery packs. The factory currently has a capacity of 35 GWh, and an actual output of approximately 24 GWh. Tesla owns the factory, but they've partnered with Panasonic who've come in to create battery cells within the factory, that Tesla then buys from Panasonic and turns into battery packs.

Recently it has seemed as if the cooperation between Tesla and Panasonic might come to end, and it's looking like Tesla will start producing its own battery cells with the help of technology acquired in the Maxwell acquisition. Elon even mentioned at this year's shareholders meeting that they might get into the mining business, to help supply themselves with the raw minerals needed to create the batteries. We should hear more details about all this during Tesla's Battery Investor Day that is planned for sometime during the second half of 2019.

You might be wondering why Tesla is going to produce their own battery cells, and is even considering getting into the mining business. I think there are three things at play here. One, I think that with the acquisition of Maxwell, Tesla believes that they can make better battery cells than Panasonic. Two, I think that Panasonic's battery cell production might be starting to become a bottleneck in Tesla's growth, because they've already admitted to converting two battery production lines that were initially meant for stationary storage, into lines that produce batteries for EVs.

Three, I think that battery cell supply and the supply of raw materials in general, is starting to be the biggest bottleneck to the growth of Tesla's EV production capacity. Tesla has unveiled three new models (Y, Semi, Roadster 2), and is about to unveil a fourth (Truck) that in terms of design all seem like they could be put into production in the next couple of years. However, the Semi is going to be delayed by at least one year (in my opinion more likely two), and I personally think the Roadster 2 won't go into production any time soon either. Furthermore, the Model 3's production has more or less remained steady at 5.000 units per week over the past 12 months, even though going from 5.000 units to 10.000 units per week should be a lot easier than going from 0 units to 5.000 units per week, which they accomplished in the first 12 months after launch. I think all of this points to Tesla currently being battery cell supply constrained, but in the words of one their VPs "We are not sitting idly by, we're making the moves required to be in charge of our own destiny, and we have solutions in place". We'll find out more later this year, during the Battery Investor Day.

Other car manufacturers on the other hand, although they are not sitting completely idly by, are definitely not in charge of their own destiny. With the exception of the Chinese car manufacturer BYD, every other car manufacturer is relying on outside suppliers for their batteries. This might not be a problem, but what if demand outpaces supply and suppliers raise their prices? Or what if a global recession hits, suppliers need to cut spending, and decide to stop expanding production for a year or two? Or what if they simply find out that it's more cost efficient for them to ramp production at a slower pace than the car manufacturers need?

It might turn out okay for the car manufacturers, but it might not. The auto industry is taking a big risk by not investing in battery technology and supply themselves. I personally rather invest in a company that is in charge of its own destiny when it comes to such an integral part of its business, than one that relies on other companies for it.

3: Vertical Integration

This brings me to my next point. Car manufacturers already outsource the production of a lot parts. With EVs and the outsourcing of battery production, the number of parts they make themselves is only going to decrease.

Sandy Munro is a widely respected automotive industry specialist. He, along with his employees, makes a living taking apart and analyzing cars in incredible detail. They then take what they learn, write reports about it, and sell those reports. Munro recently did a comparison of a number of EV vehicles, including Tesla's Model 3 and GM's Chevy Bolt. Munro reported that 70% of the Chevy Bolt is made by LG Chem (a battery, electric motor, and drivetrain supplier), meaning probably far less than 30% of it is made by GM themselves, whereas 90% of the Model 3 is made by Tesla in house.

This amount of vertical integration not only gives Tesla better margins on its vehicles than its competitors, it also gives them an unparalleled degree of control over their entire production process. The seats of the early Model S were considered one if its weak points, but since then Tesla has brought seat production in house* and significantly improved quality.

*Only manufacturer that makes its own seats as far as I'm aware

Initially it's a lot easier to just outsource things you're not competent in, but now that Tesla has learned how to build almost everything themselves, it is in my eyes a significant advantage over the rest of the automotive industry. Just like with battery production, they're in charge of their own destiny.

4: Dealerships

I guess this could be considered part of vertical integration, but I think it deserves being mentioned on its own, because there are a few very distinct advantages associated with Tesla not having to sell its cars through the usual dealership model.

Out of all established car manufacturers, Tesla is the only company that sells its cars direct to consumers. Everybody sells its cars through franchised dealerships, except for Tesla which sells all of its cars online through its website. Tesla does have on the order of 200 stores worldwide, where customers can learn about Tesla's products and get a test drive, but these are very different from ordinary car dealerships in a number of important ways, that can all be lead down to the way they're incentivized.

Tesla's stores are fully owned by Tesla and are incentivized to act in Tesla's long term interests, the success of the company, and maybe even more importantly the success of EVs as a whole. After all, Tesla's mission statement is to "accelerate the advent of sustainable transport and energy". As a result, Tesla stores educate people on the advantages of EVs over ICE vehicles first and foremost, and they create a pleasant buying experience for consumers.

Car dealerships on the other hand are separate business entities, make the vast majority of their money from servicing*. This incentivizes them to above all else make sure that customers do business with their dealership instead of another one, and to make people buy cars that have high maintenance costs. This has lead to a notoriously bad car buying experience at traditional dealerships, and to dealerships doing everything they can to stop people from buying an EV, because of their much lower maintenance and repair costs compared to ICE vehicles. Some dealerships don't even charge their EVs, so they can't be taken on test drives**.

*Tesla on the other hand is aiming to break even on car servicing. They believe that they should be incentivized to manufacture cars that need as little servicing as possible. They're also coming up with various novel ideas to improve the servicing experience, such as their mobile service fleet that comes to your house to service your car, saving you a lot of time and hassle.

** "Of the Sierra Club volunteers who were at dealerships and asked to test-drive an EV, there was not enough charge in the battery to do so 14% of the time. At Ford (21%) and Chevy (22%) dealerships, the number was even higher."

In the EV future, Tesla is setting itself up to control and add value, to everything from mining raw materials that go into their batteries to the eventual sale of the vehicle, and even the software in the car. All that traditional car manufacturers will do is basically design & assembly.

5: Superchargers

A big advantage of electric cars is that, provided you have an electricity outlet nearby where you park, you can charge your car at home, and you never have to go to the gas station again as long as you drive no more than 200-300 miles in a day. Most people rarely drive farther than that in a single day, but some people do and most people do occasionally go on road trips. To be able to do so in an electric car requires charging stations, just like we have gas stations.

Tesla realised this very early on, and started to build a network of Supercharger stations across the globe back in 2012. They've rapidly expanded this to 12,000+ chargers at 1,400+ stations in 36 countries by early 2019, allowing their customers to travel almost anywhere in Northern America, Western Europe, and large parts of Asia. Customers just tell their Tesla where they want to go, and their car will automatically calculate the most effective route, and when and where to charge using the Supercharger network. Tesla recently mentioned at their 2019 Annual Shareholders Meeting that they see a direct correlation between car sales and the availability of Superchargers. The better an area is covered by Superchargers, the more cars they sell.

Tesla by no means has a monopoly on charging stations though. As part of their Dieselgate settlement, Volkswagen has started a subsidiary that builds charging stations in the US, and plans to start building them in Canada too. There are also companies such as ChargePoint, whose entire business model revolves around building charging stations.

However, there are a few notable advantages here for Tesla. For one, as far as I'm aware the charging speed that Teslas have been able to achieve at Superchargers has always been higher than any other EV at any other charging station. This gap only just got bigger with the introduction of Supercharger version 3 earlier this year, which allows Tesla owners to charge their vehicles even faster.

Furthermore, the price that Tesla charges for electricity at their Superchargers is lower than the other charging stations. Tesla charges an average of $0.28 per KWh in the US. At Blink Network, for example, prices range from $0.39 to $0.79 per Kwh. This difference stems from the fact that Tesla does not aim to make a profit from their Superchargers. For Tesla, the Superchargers are simply a marketing tool, and an extra selling point for their car, which is in stark contrast to the other charging networks whose survival depends on making money off of the sale of the electricity.

Lastly, the Superchargers are an additional level of vertical integration, putting Tesla in charge of their own destiny. They're actively researching & developing better versions of the technology, and expanding their infrastructure as needed to support the sale of their vehicles.

It has to be said that Tesla has offered to allow any other car manufacturer to use their network of Superchargers, as long as they share equally in the costs, but so far no one has taken them up on their offer. It looks like others prefer to do their own thing, such as BMW, Daimler, Ford, Volkswagen, Audi, and Porsche, who teamed up to create IONITY, a joint venture with the mission to build a charging network in Europe. However, in its 2 years since inception, IONITY has only deployed 100 charging stations.

6: Brand

So far I've only talked about clear, quantifiable advantages that Tesla has over competitors. This next one is a little bit different, but in my eyes an advantage worth mentioning nonetheless.

Tesla's brand has been hurt a little bit as of late, for a small part due to a few missteps Elon made last year that caused bad publicity, and for a large part due to over a decade of relentless media attacks by entities that would benefit if Tesla failed (oil, automotive, short sellers. but more on this later). In spite of this, I get the impression that the majority of people do respect Elon for his genius, and for what he is trying to do to make the world a better place. I also get the impression that Tesla has a lot of positive brand awareness, especially among certain subsets of the population like millennials, younger kids, and in China.

Different people like different car brands though, and I think that in terms of sales it's hard to argue that Tesla's brand gives them more than a moderate advantage over other manufacturers at best, if any advantage at all. There is one subset of the population though with which, I believe, its brand gives Tesla a large advantage over other car manufacturers. Potential employees.

 In 2017 Tesla received a staggering 500,000 job applications, even though they only grew from about 31,000 to 37,000 employees in the same year. This means that they hired only the top 1% of job applications they received. In March of 2018 LinkedIn released a list of the top 50 companies where employees most want to work. Tesla came in at number 5, above Apple. Neither of the other large US car companies (GM, Ford) made it into the top 50.

I think this partly stems from Tesla's brand being particularly strong among younger people, and partly from the degree to which Tesla is innovating and trying to create a brighter future for humanity. Ask yourself, would you rather work at a traditional car company, that has been slow to accept the inevitable future that is EVs (many STILL don't fully embrace it), and lobbies against EV regulations in countries like China, where basically all manufacturers except for Tesla have lobbied to slow down the government's EV regulations. Or would you rather work at Tesla, the company that has almost single-handedly changed an entire industry, is at the cutting-edge of these new technologies, and where you can be part of one of the most exciting things happening in the world today? Looking at the number of people that apply to work at Tesla, I think it's clear that most people's answer to that question is that they'd rather work at Tesla than at any of the other car companies.

7: Tesla's Customers

Tesla's customers LOVE Tesla. Have you ever heard of people going to work for a company for no salary, no benefits, no compensation whatsoever? Me neither, except for when it comes to Tesla's customers who volunteered to help to deliver cars during the Model 3 ramp at the end of Q3 2018. During this quarter Tesla had to deliver approximately three times as many cars in the US as they had ever attempted to deliver before. Somebody tweeted to Elon asking if there was any way he could help with the end of quarter delivery nightmare Tesla was facing. Elon tweeted back that any current owners who were willing to help should show up at delivery centers. Hundreds of them did, and voluntarily helped the company they love through a tough time.

Elon and Tesla have attracted what some consider to be a cult-like following, but whether you believe that there is a 'Tesla cult' or not, it is hard to argue with hard numbers that show that Tesla's customers love its products:
These are absolutely amazing numbers. I don't personally own a Tesla, because I have no driver's license, and I think that for the foreseeable future the stock is a better investment than any of their cars, so I can't say I know the ins and outs of the Tesla community. What I do know however, is that although Tesla is not perfect, and there seem to be some structural problems in terms of service, customer support, and overall customer experience, the overwhelming majority of its customers LOVE the cars.

8: The Cars & Demand

As a company, having a loyal customer base that loves and supports you is fantastic. But I think almost anybody could create a company, make a shitty product, and get at least one supportive customer. It's equally important that your products are great and generate a lot of demand, so that your loyal customer base includes more people than just your mom.

The most objectively amazing feature about teslas, and one of their main design principles is the cars' safety. In 2013 the National Highway Traffic Safety Administration (NHTSA) tested the Model S as the safest car they had ever tested. In 2017 the same NHTSA tested the Model X, and awarded it 5-star ratings in all categories, the only SUV to have ever received such a rating. In fact, the Model X had the lowest probability of injury of any car the NHTSA had ever tested, second only to the Model S. This remained unchanged until they tested the Model 3 in 2018, which then became the safest car the NHTSA ever tested. All in all it seems pretty clear that Tesla makes the safest cars in the world.

Teslas also have a few weak points. There have been mentions of panel gaps in some cases, and overall I get the impression that Tesla's build quality, albeit not bad, isn't quite up to the level of some of the large car manufacturers just yet. It also appears that there is a subset of the population that feels like Tesla's interior isn't as good as that of some of the luxury car manufacturers, like BMW and Mercedes. Nonetheless, the most common thing I hear on the internet from people who have driven the Model 3 is: "After driving the Model 3, your expectation of what a car is supposed to be is changed forever", so it seems as if Tesla is making pretty damn good cars. As an investor, however, I can't rely on sentiment alone, so let's get back to taking a look at some hard numbers, to see what kind of demand there is for Tesla's cars.

2015 Large Luxury Car Sales US:

 In 2015, three years after it was launched, the Model S was absolutely crushing the competition, and clearly taking away significant market share from all competitors in its segment. 

December 2018 Large Luxury Car Sales US:

In 2018, the Model S had a nearly 40% market share in its segment in the US, selling twice as many cars as its nearest competitor, the Mercedes S-Class.

December 2018 Large Luxury SUV Sales US*:


*For some reason, the Model X is listed in the Medium Luxury SUV Class on the site where I found these sales data, but it is twice the price of the vehicles in that class, and should be part of this Large Luxury SUV Class.

The Model X is not dominating its class like the Model S is, but it's still posting solid sales numbers. Furthermore, I suspect that the demand for Model X is actually higher than this. The SUV market is much larger than the market for sedans, so I suspect that Tesla simply stopped expanding S+X production in Q3'16 when they first produced 25,000 S+X in a quarter, and moved all its focus towards ramping up Model 3 production. I don't think anybody could argue that if they had opened more stores in other parts of the world (SEA, S-America, Africa, India, East-Europe), they could've definitely sold more S+X, but I think they simply had to shift focus towards the Model 3... for now.

I could've perhaps turned this into its own entire segment, although it'd be a bit weird to say that not being in certain markets is an advantage to Tesla. The fact remains however, that Tesla has not yet established a presence in large parts of the world, but undoubtedly will in the future further increasing demand. Even in China, Tesla may have established a presence, but mostly due to tariffs they haven't reached anywhere near their full potential yet in the biggest car market on the planet.

December 2018 Mid-size Luxury Car Sales US:


Now onto perhaps the most impressive, and astonishing numbers of all. First of all I want you to just look at those numbers, take them in, and be amazed at the ridiculousness of what are the 2018 Tesla Model 3 sales numbers. Second of all, I need to make some footnotes, some making these numbers even more impressive, some making them less impressive:
  • The 25,250 Tesla Model 3s delivered in December of 2018 were part of an end of quarter (EoQ) delivery push. Looking at other manufacturers' numbers it looks like other manufacturers might deliver more cars in the last month of a quarter as well, but Tesla probably delivered 63,150 Model 3s in the US in the entire fourth quarter, so their average monthly deliveries were approximately 21,000.
  • As for the 140,317 Model 3s that were delivered in the US in all of 2018, it has to be said that Tesla only really fully ramped production during the second half of the year. They delivered ~25,000 Model 3s in the US during the first half of 2018, so that means they delivered ~115,000 Model 3s in the second half, which would come out to 230,000 Model 3s in a full year of ramped up production.
  • There were some other factors in play that caused Model 3 deliveries to be extremely high in the second half of 2018, one of which is that Tesla had a lot of pent-up demand for the Model 3 in the form of reservations.

All in all, I think it is far too early to make conclusions about long-term Model 3 demand, as they've only just started overseas deliveries, and have mentioned currently being battery cell production constrained. They also haven't started production at Gigafactory 3 in China yet, which will lower prices significantly, and is slated to begin towards the end of this year. But I think the signs so far are amazing. Besides the sales figures I already showed, 63% of trade-ins are not luxury cars like the Model 3, but far cheaper cars such as the Honda Civic and Toyota Prius. This shows that people who normally buy much cheaper cars (a far bigger market), are willing to spend much more money than they normally would to buy a Model 3. So even though the Model 3 is still somewhat expensive, it's going to take away market share from a far larger market than people are expecting. Last but not least, the Model 3 has been the #1 best selling car by revenue in the US over the past 12 months.

Before I can stop talking about demand, I have to mention that Q1'19 was actually a very weak quarter for Tesla:

March 2019 Large Luxury Car Sales US:


March 2019 Large Luxury SUV Sales US:


March 2019 Mid-size Luxury Car Sales US:

The S+X deliveries dropped by nearly 50% YoY, and Q1'19 Model 3 deliveries dropped by nearly two thirds compared to Q4'18. I don't think there's any reason to be concerned about the Model 3. Tesla produced more Model 3s in Q1'19 than in Q4'18, in spite of Q1 being seasonally weaker than Q4 in the auto industry. These numbers are low simply because they started shipping Model 3s outside of the US for the first time ever, and experienced temporary logistics difficulties. Besides, even in spite of all of this, the Model 3 is STILL outselling its nearest competitor 2-to-1 in the US, but I suspect long term demand to be much higher than this.

For the S+X however, we are comparing Q1'19 to Q1'18 numbers, so there is no seasonality involved, and they have been delivering outside of the US for years. So what is going on here exactly? At first I was a little worried about these numbers, but during their Q1'19 earnings call Tesla explained that they were retooling their S+X production lines, to prepare for the Raven update that they released in Q2, and therefore couldn't produce more cars than this. A few other things could impact S+X demand negatively in the short-term like the continued phasing out of the US EV tax credit, the rumours of a big S+X refresh, and Model 3 cannibalizing sales. It's something to keep an eye on in the short term, but I don't think this is something that's going to impact Tesla in the long term.

*I've used US Sales numbers throughout this section, because they are the most easily accessible. It's possible that Tesla, on average, has a small advantage in the US compared to other markets, because of patriotism (although plenty Americans hate Tesla), import duties that I'm unaware of, or some other reason. We can't blindly extrapolate these numbers across the world and expect them to be the same everywhere. Tesla definitely has not performed well in certain markets such as Germany, where the population is very protective of its auto-industry (Mercedes, Daimler). The same is true for Japan, where it's more common to not be able to charge an EV at home, and the Model S is too big to enter many parking garages. However, I still think one can look at these numbers and conclude that Tesla knows how to make extremely desirable cars, and anticipate that Tesla will over time be able to capture large market shares across the world with their current, and upcoming vehicles.

Automotive Advantages Summary

And with that, we've come to the end of the list of advantages that Tesla has over other car companies. To sum this all up, let's look at this list from a very high level overview.

I think Tesla's vertical integration will lead to better margins and lower prices, but it's not like it stops competitors from making compelling EVs. Dealerships cause disadvantages to traditional car manufacturers right now, but if they put effort towards it, they could come up with ways to minimize these. Tesla's loving customers and Superchargers are both strong assets, that are not going to go away anytime soon.

I have a feeling not too many people agree with me about Tesla's brand giving Tesla a significant advantage in terms of attracting talent, and that this makes a huge difference. It is impossible to prove or quantify, but Tesla has been executing extraordinarily well since the launch of the Model S, with a revenue Compounded Annual Growth Rate (CAGR) of 61% since 2013*. I think that the quality of Tesla's talented employees is a big deal, and an even bigger asset than its customers.

*Much higher if you count 2012, which many bulls do, but that is unfair in my opinion. 2013 was their first full year of Model S production.

But in my opinion, far and away Tesla's biggest advantage is its lead in batteries. I think it's very obvious that Tesla makes the best EV batteries in the world, and that is before integrating Maxwell's technologies. In terms of battery supply I'm less worried, but still worried for other manufacturers, that for the scaling of production of the most important part of their future vehicles they will be reliant on other companies.

All of these advantages are clearly reflected in the demand for Tesla's vehicles. It's hard to tell exactly how high that demand is at the moment, because they don't operate in all markets yet, and they've thus far always been supply constrained. It is clear however, that Tesla has established itself as a car manufacturer that has to be taken extremely seriously by its competitors.

Tesla Energy


When bulls talk about Tesla, they sometimes mention that they believe Tesla is undervalued because it's being valued as a car company, even though it is much more than that. Bears would argue that Tesla is overvalued regardless of what kind of company you value it as. During the first half of this decade I would've agreed with the bears that, if you take future potential out of the equation, Tesla was overvalued regardless of what kind of company you valued it as. Nowadays this is probably not the case anymore, but I digress.

Although Tesla's main business is undoubtedly automotive, and I don't see the energy side of Tesla's business adding substantial value for the next decade, I do think it's worth going over Tesla's Energy business, and understanding what it is, and how it works.

Powerwall & Powerpack

Tesla first unveiled the Powerwall and Powerpack in the first half of 2015. The concept behind the products is fairly simple. If you don't believe in global warming, we will have to transition to sustainable energy sometime between now and when we run out fossil fuels. If you do believe in global warming, we will have to transition to sustainable energy sometime between now and as soon as possible. Either way, transition to sustainable energy generation and consumption we must.

Tesla's cars (and other EVs) are one part of this solution, as they make a large part of our energy consumption (transportation over roads) sustainable. Powerwall and Powerpack solve a major problem of the other part of this solution. The problem with the other part (sustainable energy generation) is that it is not stable. Solar power depends on the sun, which doesn't shine equally strong 24/7/365. Wind power relies on the wind, which fluctuates in power constantly. Powerwall and Powerpack solve this problem by storing some energy in batteries during times of peak energy generation, so that it can be used when energy generation is insufficient to cover the demand for energy on its own.

It makes a lot of sense for Tesla to leverage its competency in batteries to create energy storage products, because as we've established, Tesla is the world leader when it comes to EV battery tech. Stationary storage batteries might not be the exact same thing, but they are very similar, and I think Tesla will be able to leverage their EV battery tech expertise to also make extremely compelling stationary storage products.

So far Tesla has made some bold claims about their energy division. They've said that they expect it to be more valuable than their car business long term. And they've said that for the foreseeable future they expect to each year install as much energy storage as they have in all previous years combined. Thus far, they have delivered. They installed 98 MWh in 2016, 410 MWh in 2017, 1,040 MWh in 2018, and are guiding for between 2,000 and 3,000 MWh in 2019.

So far so good, right? Well, there are a few reasons why I'm pretty neutral when it comes to Tesla's stationary storage business:
  1. Nobody outside of Tesla knows what kind of margins Tesla is getting on their stationary storage products. We know the margin of Tesla's overall energy business (which was not amazing in 2018 at 12%), but it includes energy generation (more on this soon) in addition to the energy storage products.
  2. For the time being, it is very small in comparison to Tesla's automotive business. Tesla's 2018 automotive revenue was 18.5B$ in comparison to 1.5B$ for its energy business, which again includes energy generation.
  3. Most importantly, Tesla is currently battery cell supply constrained. If they had an endless supply of battery cells, I'd be excited to see them grow their energy storage business, but I never want to see them take a battery that could've been put into a car, and put it into a Powerpack or Powerwall instead. This is because every single battery Tesla puts inside one of their cars, will pay off far more when Tesla solves autonomy and is used to power a robotaxi (more on this later).
I do think stationary energy storage is extremely important in transitioning our world to sustainable energy, and that it can eventually be a great part of Tesla's business. However, in light of autonomy I want Tesla to focus all their efforts on their automotive business, until they have a huge fleet of robotaxis on the road. After that I'll fully support them in growing their stationary storage business.

SolarCity & Solar Roof

Tesla's energy generation business started with the acquisition of SolarCity in late 2016. SolarCity was founded in 2006 by Elon Musk's cousins. Elon provided 10M$ in startup capital, and became the chairman of the board. Their business model was to be the equivalent in the solar business of what Dell is in the computer business. They didn't build any of the solar panels themselves, but instead focused on sales, installation, and repairs.

In 2008 they also started to give their customers the option to lease the panels. This means that SolarCity paid for all of the up-front costs, and the customer simply got a reduction in their monthly bill, because the cost to lease solar panels is less than the cost of electricity. By 2016 this "no-money-down solar" business model became the most popular in the USA, and SolarCity had become one of the biggest solar installers in the country. It also meant however, that SolarCity had accumulated a lot of debt, because it had had to pay for all the up-front costs associated with solar leasing.

In June of 2016 Tesla announced that it would be acquiring SolarCity pending shareholder approval. Elon said at the time that it was a 'no-brainer', because it would be able to seamlessly integrate SolarCity's solar products with its own battery storage products that it had recently launched, and that there were a lot of synergies between the companies. Although I understood the big picture possibilities at the time, I was hesitant. I was extremely optimistic about Tesla as a company, and understood both the risks and the opportunities, but my knowledge about SolarCity was very limited. I ended up not being able to cast my vote on the acquisition, because my stocks were with an overseas broker, but I would've voted in favor of it, just because of the level of trust Elon had built up with me.

If I could go back in time and vote on the merger now though, I would definitely vote against it. Tesla has stopped the "no-money-down solar" leasing, which was probably a good decision in order to stop accumulating debt and improve its balance sheet. However, the debt that SolarCity built up from before the acquisition still remains, and it also means that a lot of what was SolarCity's business has simply disappeared.

The only thing that Tesla might've gained from the acquisition is intellectual property, but the upcoming Solar Roof to me seems very different from ordinary solar panels, and SolarCity never created any solar panels in the first place (although it was planning on doing so in cooperation with Panasonic). So I question if Tesla really needed to spend nearly 3B$ in stock to acquire SolarCity to create this new Solar Roof product. I don't know for certain, but I think Elon and Tesla most likely did not do the right thing here. However, let's focus on the future, and let's see what this 3B$ in stock bought us.

Solar Roof is exactly what it sounds like. It's a roof that is also a giant solar panel. The good things are that it looks like the most beautiful roof you've ever seen, it apparently will last longer and be stronger than a normal roof, and it is an extremely elegant solution for residential solar, that could play an important part in solving the problem of sustainable energy generation. The bad things are that it is looking like it'll have very high up front costs, and as an investor I have absolutely no idea yet what the financials will look like.

To me the energy generation part of Tesla's business is about even, maybe slightly worse than the energy storage part. On one hand it doesn't consume any of Tesla's precious battery cells, but on the other hand it's an unproven business, and the technology is different from their bread and butter EV technology. So yeah, it has significant long term big-picture potential, but it's not going to contribute to Tesla in any significant way any time soon.

Summary of Tesla Energy

All in all, I think one's view of Tesla Energy will be directly correlated to one's degree of general optimism. If you put on a bear hat, you could say that it's nothing but a distraction from what is really important. If you put on a bull hat, you could say that both storage and generation are enormous opportunities, especially because the energy storage market is projected to grow exponentially to 158 GWh by 2024. In my opinion it's somewhere in the middle. Yes, Tesla Energy has huge potential, but it's going to be small for quite a while compared to Tesla's automotive business, on which they need to be laser focussed for the next 5-10 years. Tesla's success is entirely dependent on its automotive business. If that fails, Tesla Energy won't save Tesla. If it succeeds, Tesla Energy can eventually make Tesla even more successful.


Tesla Investor Loyalty Program


The things in this next section aren't things I'm excited about daily, nor do I think they add much if any value to current stock price. You could say they're part of a Tesla loyalty program of sorts, where you get lottery tickets to win prizes. You probably won't win anything, and even if you do, it's probably a small prize. Winning a huge prize in a lottery is highly unlikely after all. But if turning highly unlikely things into reality required a lot of luck, Elon might be the luckiest man on earth, so let's take a look at what I call the 'Tesla Investor Loyalty Program'.

Tesla Insurance

Tesla Insurance is already right around the corner. Elon mentioned at the Annual Shareholders Meeting, that they will launch it as soon as they complete a certain acquisition. I think the value proposition of Tesla Insurance today is that, although Tesla cars are statistically the safest cars according to the NHTSA, their insurance premiums are disproportionately high. Tesla aims to change this by launching their own insurance program.

However, the bigger value of Tesla Insurance lies in Tesla's plans to launch a robotaxi service called Tesla Network. If Tesla manages to create self-driving cars, their software will be driving the cars and they will be liable for any damages. This will require them to have a lot of insurance, so it makes sense to integrate this into their business and do it themselves.

In-Car Entertainment

Tesla is already adding more and more fun features to its cars in the form of games and Easter eggs. As of right now they're no more than that, fun additions to the car that might be an extra selling point to certain customers. When Tesla manages to solve Autonomy however, there is quite a bit of opportunity to turn this into a big business.

Tesla already has a lot of screen real estate in its current cars compared to most of its competitors, but I suspect that Tesla's designers are already thinking about what their cars will look like after autonomous vehicles (AVs) become mainstream. I suspect that Tesla's upcoming pick-up truck, as 'cyberpunk' and different as Tesla has teased it to be, might be nothing compared to the next car they will reveal after that. Tesla's future AVs might end up becoming something in between a cell phone and a movie cinema. And just like Facebook has created a business around the consumer attention it captures, there could be possibilities for Tesla to do something similar in the future.

Tesla Home

Like I mentioned in the Tesla Energy section, I'm pretty neutral on Tesla Energy in the near term, but in the long term Tesla could turn Solar Roof and Powerwall into huge businesses. If I put on a bullish hat and look at Tesla's Solar Roof + Powerwall 2 unveiling, I could see them providing a large number of people with sustainable energy generation, storage, and also consumption in the form of EVs.

I think it's not too far fetched for Tesla to create electrical appliances that can be integrated into a smart home alongside their Solar Roof and Powerwall. As a matter of fact, Elon has already mentioned wanting to do something with air conditioning on a few occasions, so he's clearly thought about this. Also, looking at their degree of vertical integration it doesn't seem like Tesla is afraid to get its hands dirty, and design and manufacture a wide variety of things from the ground up.

Currently Tesla doesn't have enough resources to spread itself this thin, but if at some point they're swimming in cash like Apple, I wouldn't be surprised to see them get into this.

Electric Airplanes

One of the improvements that Maxwell has made in battery tech is an improvement in battery energy density. Tesla's current batteries appear to have an energy density of 250 Wh/kg. Maxwell has demonstrated energy densities of over 300 Wh/kg, and has identified a path to a battery energy density of >500 Wh/kg. This means that if all goes according to plan, in the future the same battery will be able to hold more than twice as much energy.

When I first heard about this, I instantly thought of Elon's plans to design an EVTOL (Electric Verticle Take Off and Landing) Aircraft. Elon has talked about having a design for such an aircraft numerous times, and said he would love to build it, but that he believes battery energy density has to improve to over 400-500 Wh/kg to make it work.

It's definitely a long term thing, but I think that if Tesla succeeds in their automotive business, the chance that they announce plans to produce an electric aircraft by 2035 might be as high as 80-90%. After all, their CEO also just happens to be the lead designer at the world's number one rocket company, so he knows a thing or two about aerospace engineering.

The Boring Company

The Boring Company (TBC) is a small company started by Elon and a couple of SpaceX employees, with the goal to improve tunneling technology. They hope that with improved tunneling technology (lower prices, faster speeds, etc) they'll be able to build high-speed, underground transportation systems.

TBC and Tesla are already cooperating in a couple of ways. TBC is using Tesla's EV technology to dig tunnels, and TBC planned on using electric cars and sleds from Tesla to traverse its tunnels. It currently seems like they've shifted to simply create underground roads, and to rely on autonomous software to drive through them.

Like the others, this is a long shot and relies on TBC being successful, but especially if they end up deciding to rely on AV tech to navigate through tunnels, I could see it being a lot easier for Tesla to cooperate with TBC and make their cars work in the tunnels, than it would be for other companies.

The TILP

To summarize, Tesla has a few very cool potential future opportunities. None of these significantly influence my opinion on what Tesla is worth today, but Autonomy was in a similar situation until very recently, and now it's the most important reason why I'm excited about Tesla. So who knows.

Autonomy


I used to think that if things went well, I could make 5-10x, maybe even 20x by 2030 on my 2015 investment in Tesla. I still think this is true, but in addition to that I now think there is a real possibility of seeing 50-200x returns by 2035-2040, maybe even a lot more. Some of the things I'm going to say later on in this post will sound crazy, but after some research, thinking, and financial modelling, I think the bull case for Tesla in terms of autonomy is legitimately just absurd. We'll get to that when we take a look at financial models later on. This section is all about autonomy, and why I think Tesla has a large lead in the race to get to full autonomy.

A brief history of Autopilot

Tesla first released Autopilot towards the end of 2015 with the help of Mobileye. Mobileye has spent the last two decades providing the automotive industry with vision-based Advanced Driver-Assistance Systems (ADAS), that make driving a car slightly easier through features such as forward collision warning systems, and lane departure warning systems. Mobileye provided Tesla with a camera, a chip, and probably some software too, that allowed Tesla to launch its first version of Autopilot.

In July of 2016 Tesla ended its partnership with Mobileye, and a few months later it announced its plans for developing Full Self-Driving (FSD) vehicles. Tesla claimed that every car built from October 2016, would have all the hardware needed for FSD, and that it was just a matter of time before the software would be able to drive the cars by itself. The only thing that might need to be upgraded, the computer chip, they said was easily replaceable and would be upgraded for free if necessary. Elon also claimed they'd be able to do a coast-to-coast drive by the end of 2017.

However, the end of 2017 came and went without a coast-to-coast drive. During 2017 and 2018 very little happened in terms of Autopilot progress actually. Maybe this is why most people seem to have missed what has happened so far in 2019 (Navigate on Autopilot (NoA) ), and maybe this is why nobody seemed to believe much of what Elon said during Tesla's Autonomy Investor Day this April, where Tesla showed the world their lead in self-driving technology. But before we take a look at where Tesla is, let's first take a look at the competition.

The Competition

  1. Waymo. Waymo has been around for over a decade. It started in 2009 at Alphabet (back then called Google) as Google's Self-Driving Car Project, but has since been spun off into an Alphabet subsidiary. They make their sensor suite in house, which consists of 360 degree cameras, radar, and LIDAR. Their sensor suite is the most extensive one in the business, and costs a little under $10,000. They have a few hundred test cars are on the road, and have launched an actual robotaxi service in Phoenix, Arizona, but there are still safety drivers behind the wheel to take over when necessary. Waymo has been privately valued at over 100B$.
  2. GM's Cruise. Cruise Automation started in 2013 as a company developing kits to retrofit vehicles with limited self-driving capabilities. In 2015, they changed their strategy and began to develop software for fully self-driving vehicles. GM came into the picture in March of 2016, when they acquired the startup for an undisclosed amount rumoured to be anywhere from 500M$ to over 1B$. Similar to Waymo they use a sensor suite of cameras, radar, and LIDAR, and they plan to launch a commercial service as soon as 2019, but recently reports have come out stating that they are very far behind their targets, and they've been burning through a lot of money. Most recently GM's Cruise was valued at 19B$.
  3. Uber. Uber has a self-driving car program, but that about sums it up. Although it was reportedly valued at over 7B$ after receiving an investment from Toyota, it's rumoured to be burning 20M$ per month, it's been involved in numerous high-profile crashes, and they themselves have gone on the record saying "We expect it's going to take a long time to have this technology at scale". I don't think anybody following the space would argue for Uber being one of the leaders.
  4. Apple's Project Titan. The Wikipedia page for Apple's self-driving electric car project, code-named Project Titan, is an interesting read. There are rumours about Apple developing an Autonomous Electric Vehicle (AEV) dating back to 2014. In 2016 Elon Musk said that "It's obvious Apple is working on an AEV. It's pretty hard to hide something if you hire over a thousand engineers to do it". Since then it seems as if Apple has dropped its plans to develop an EV, but in 2017 CEO Tim Cook acknowledged publicly that they are working on FSD tech. Although details are scarce, it looks like they have a little under 100 cars on the road for testing, and around 5,000 employees working on the project, so they're definitely taking it very serious.
  5. Mobileye. Tesla's old partner Mobileye was acquired by Intel in 2017 for 15B$. Mobileye has always heavily relied on computer vision and cameras to make their products work. In theory this sets them up to be a serious contender in the self-driving tech space. They're currently working on developing FSD tech with a sensor suite that includes radar and LIDAR in addition to cameras, and they plan to launch a robotaxi service in Israel, where they are based, by 2022.
  6. Nissan ProPILOT. Nissan has an ADAS in some of their vehicles called ProPILOT. If customers pay for the system, they get an ADAS that helps them in single lane highway driving conditions. They have a 4-step road map to expand this to multi-lane highways, city streets, and eventually turn it into a FSD system by 2022. Their current sensor suite only has a forward-facing camera, and a forward-facing radar though, which is definitely not enough to go completely autonomous.
  7. Others. If I were to name all of the companies in this space, I suspect the technology would be finished sooner than I'd finish that list. Daimler just announced a partnership with Bosch, pretty much every car manufacturer is doing at least something, and there are dozens of startups trying to be the first to create a fully autonomous vehicle. There are also billions of dollars being invested by automakers, and companies like SoftBank and Amazon.

Tesla's Autonomy Strategy

If you want to fully understand in detail how Tesla plans to solve autonomy, I highly recommend checking out their Autonomy Investor Day on YouTube. It is two and a half hours long though, and most of it is extremely technical. As such, I will do my best to sum up Tesla's autonomy strategy here.

The most fundamental underlying belief that supports Tesla's autonomy strategy is that just like humans drive cars using their eyes for vision, computers can learn to drive cars using cameras for vision. Tesla's sensor suite consists of 8 cameras placed around the car, and additionally uses 360 degree short range sonars, and a single forward-facing radar to help detect objects in the direction in which its cars travel the fastest. They use images, video, and other data from their sensors to train a neural network to detect all the objects relevant for driving. This includes other cars, lane markings, traffic lights, pedestrians, curbs, and many more things.

On top of their powerful neural vision net Tesla builds various safety features, and software that actually drives the car. An example of a safety feature is Automatic Emergency Braking (AEB). Examples of software that actually drive the car are Summon, which allows you to summon your car out of a parking spot or parking garage, and Navigate-on-Autopilot, which can navigate highways from on-ramp to off-ramp including lane changes. Tesla's plan is to continually add new software to the point where eventually their cars can navigate all types of roads. At that point it will just be a matter of increasing the safety until the software makes a mistake less frequently than a human driver does.

When they've built software that is capable of driving safer than a human, they'll have to gather enough data to convince regulators that their cars are safer than human drivers. After receiving regulatory approval Tesla plans to launch their robotaxi service, called Tesla Network, to compete with Uber and Lyft. Tesla will not allow any of their self-driving cars to be used on other ride sharing networks, but owners can choose to add their car to the Tesla Network, where it can earn money driving people around. Tesla will take a percentage of the profits, and also operate their own fleet of cars to supplement their customers' cars.

Why Tesla will dominate autonomy

All of this is a solid strategy, but most of this is pretty similar to their competitors. So why do I think Tesla is going to dominate the future robotaxi market? They have a number of advantages, some of which appear to be very hard to overcome for competitors.

  1. Data. When it comes to AI and training neural networks, data is king. In terms of self-driving, the more data you have, the better your neural net will be at correctly identifying objects. The more data you have, the better your software will be at detecting other vehicle cut-ins. The more data you have, the more crazy edge cases you will be able to teach your AI about. Basically, the more data, the better.

    When it comes to data, Tesla isn't just ahead, they're not even just dominating, they're so far ahead that it seems like nobody else is even trying. Most competitors have about a hundred (or maybe in the case of Waymo a few hundred) cars on the road with expensive LIDAR equipment to test FSD software and gather data. Mobileye has their cameras in a few million cars, but these are cars sold by its customers such as BMW, so I don't believe they have any way to collect this data, and it's only from one camera, which is not enough information to support FSD. Nissan's ProPilot is used by about 350,000 cars, but like Mobileye this is only a single camera plus radar, and I am unsure if Nissan has any way to gather this data. Furthermore, Nissan only installed the camera on the cars of customers who paid for the feature. This shows a clear lack of understanding of the importance of data, and autonomy as a whole.

    Tesla, understanding the utmost importance of data when it comes to achieving full autonomy, has installed their entire sensor suite on every single car sold to customers since October of 2016, regardless of whether the customer paid for Autopilot or not. As a result, Tesla is currently gathering data from 500,000 customer cars every single day. In a year from now this will be approximately 1,000,000 cars, and in a few years they will be gathering data from millions of cars. Basically, their advantage in terms of data is only going to increase from here... exponentially.

    Data is not just paramount for creating a FSD car. Remember that I mentioned that part of launching a robotaxi service consists of convincing regulators that one's cars are safer than a human? The     average miles per accident in the US is 165,000 miles, and Tesla's lead competitor Waymo is driving the second most autonomous miles. If they developed a car five times as safe as a human today, went to regulators to request regulatory approval a year from now, and showed them they only had three accidents over (let's be generous and say) 3,000,000 miles. Regulators would probably laugh in their face and say they could've gotten lucky.



    Tesla's fleet, on the other hand, drove 20 million miles per day in September of 2018. They had approximately 400-450k cars on the road back then, including some cars without the full sensor suite, but today Tesla has about 500k cars on the road with the full sensor suite, so they are gathering more than 20M miles of data every single day.     Tesla is gathering more data in a single day, than all its competitors combined in an entire year, and this difference is only getting bigger.

    If another car manufacturer is smart enough, it could replicate this model, but the others simply don't stand a chance. None of the non-car manufacturer competitors can afford to buy this many cars to put on the roads and gather data. Neither can Tesla, but Tesla's customers are paying Tesla to build the cars they need to gather all the data. Simply genius. If I could only name one reason why I'm bullish on Tesla, this would be it right here.
  2. Test Drivers. But data gathering isn't the only thing that Tesla's customers are helping Tesla with in terms of autonomy. For safety reasons Tesla's competitors need to employ people to sit behind the wheel of the cars in their test fleet. Tesla's customers are literally paying Tesla to do this job.

    Part of how Tesla makes its FSD software better and better, is looking at ''Autopilot interventions". Whenever a customer is driving a Tesla in autopilot mode and notices an unsafe situation, they will take over control of the vehicle to make sure an accident doesn't happen. Every time this happens, the car sends a notification to Tesla's Autopilot Headquarters along with all the data surrounding the event. Tesla's Autopilot team then looks at what happened, and can improve the software so that it doesn't happen again. If a car manufacturer is smart, they could do the same thing, but there is no way any of the other competitors can hire half a million test drivers, whereas Tesla's customers are paying Tesla to be their test drivers. Again, simply genius.

  3. Hardware Chip. In 2016 Tesla poached a team from AMD consisting of some of the top chip manufacturers in the world, to build a custom chip for Tesla's Autopilot team. To power their software, Tesla needed a very powerful computer that was cheap, and most importantly did not consume too much energy, because that would negatively impact range.

    The team set out to work, and three years later in April of 2019 the new computer chip went into production. During Tesla's Autonomy Investor Day, the lead designer compared the new chip to the old chip, and noted that it is 20% cheaper, similar in terms of power consumption, and 2100% more powerful. This chip, he explained, was designed especially for Tesla's needs, and Tesla's needs alone. Furthermore, they are already halfway towards finishing an even better chip, that should go into production sometime early in the next decade.

    There were a lot of technical details in the presentation that I did not understand, but I understood the most important part. Tesla has built a chip that, when it comes to FSD, is a number of years ahead of the competition, and is working on another chip that is even further ahead.

    Now I must say that FSD software may not be ready yet in the next few years, and by the time the software is ready, competitors might also have access to a similar chip capable of full autonomy. However, even when a self-driving car is already safer than human drivers, it will always be important to continue to increase safety. So although this part of Tesla's lead will not stop competitors from creating a FSD car at the same time, or even faster than Tesla, it will allow Tesla to create more powerful FSD software, which will lead to an advantage in terms of safety over its competitors.

  4. Vertical Integration. This is not as big an advantage as the other three, but the degree of control Tesla will have over everything in the AEV robotaxi future is astonishing. They're involved with everything from the battery tech that determines an EV's efficiency, and therefore the cost to operate a robotaxi, to the charging infrastructure needed to support robotaxis*, to the manufacturing of the cars themselves, to the hardware powering the software, to the software itself.

    *Tesla already has a design for a robosnake, self-charging EV charger.

  5. LIDAR. Almost all companies working on FSD cars use LIDAR as an important part of their sensor suite. LIDAR is a sensor that shoots lasers through the air to detect objects. The advantages are that in most conditions LIDAR is highly accurate, and it doesn't need tons of data to get to that point unlike a vision-based system.

    Elon has refused to use LIDAR at Tesla, because he believes it is an "unnecessary, expensive crutch". LIDAR is very expensive*, and does not work well in heavy weather conditions. Elon believes that LIDAR will get you 99.9% of the way to autonomy very quickly, but that a good vision-based system is a necessity to get to 100%, and that LIDAR is not essential.

    *    Sometimes as much as 75,000$, although Waymo reduced that by 90% by designing and building their own LIDAR system from the ground up.

    Elon's opinion has been very controversial, and almost every competitor (including all companies I listed above) use a system that includes LIDAR. In 2016 Anthony Levandowski, a former Google engineer, even said "we have got to start calling Elon on his shit". However, that same engineer has since started a self-driving truck startup under the name Pronto, and opted not to use LIDAR. Researchers at Cornell have also     backed up Elon's claims, saying that a vision-based system can do anything that LIDAR can. And Lex Fridman, an industry expert, has agreed with Elon that cameras are the way to go. He has also mentioned that he's worried about how far ahead Tesla is, because competition is a good thing. I also think that although Elon isn't an objective outsider, he is one of the world's top engineers, and a very smart person in general, so his opinion should also be given a lot of respect when it comes to technology. Time will tell for sure, but as of right now I believe that Tesla's approach is correct, and that the competition is wasting time working on LIDAR.

  6. Scalibility. Tesla is also one of the few, if not the only company that is not geofencing their FSD technology. Tesla's self-driving technology works anywhere in the world regardless of location. Most other companies, such as Waymo and Mobileye, are launching their services in specific parts of the world, and are using high-precision maps to give them an advantage. These high-precision maps, like LIDAR, can give the impression of a lot of progress in a short amount of time. But it's not easy to scale and make a high-precision map of every single road, and if something about the road suddenly changes, the high-precision map suddenly works against you, because the road is different from what the car thinks it is. Tesla said they have tried to previously use high-precision maps, but quickly reversed course, because they realised it did more harm than good.

    So in short, Tesla's solution works everywhere, whereas competitors' solutions work much better in some locations than others. A pretty big deal.
Out of these six advantages I think vertical integration is a minor advantage. More of a nice to have, than something that will win Tesla the race, or help them achieve total dominance. The LIDAR and scalability issues of competitors could be solved overnight with a strategy change. But Tesla has been working on what I see as the correct strategy for a lot longer, so they have a big lead, and every day that Tesla's competitors continue to spend time and resources on dead-ends, is a day that Tesla increases that lead.

The computer chip is an advantage that I don't think anyone can argue about. Tesla has a clear, quantifiable lead here, and although it won't stop a competitor from creating a FSD car, it is a piece of the autonomy puzzle. It gives Tesla an advantage in computing power and therefore safety, at the very least.

Tesla's strategy of turning their customers into investors* and test drivers for their FSD development program is what I think is going to win them the race to the first fully self-driving vehicle. Companies like Waymo and Apple literally can't compete, unless they partner with a car manufacturer and replicate Tesla's model. Car manufacturers with AV programs can theoretically copy what Tesla is doing, but if Nissan and their ProPILOT is anything to go by, they simply lack the understanding needed to see what needs to be done. They're also notoriously bad at software, and solving autonomy is mostly a software problem.

*A very nice analogy in my opinion. Not only do their purchases help Tesla develop AVs, if Tesla is successful, they will see a very nice return on investment, because a self-driving car is many times more valuable than a non-self-driving car.

To summarize, I think Tesla is the clear leader in the race to create the world's first EAV, and unless competitors are going to wake up and change their strategies soon, I think Tesla is going to finish first by a large margin. Furthermore, if Tesla finishes first, depending on their production rate at the time and how long it takes the second place finisher to finish, I think it could be hard for anybody to ever catch up to Tesla. I see an unlikely, but possible scenario, wherein Tesla would end up with a monopoly on transportation over roads.

Let's assume that if regulators are shown a large enough data set showing that a company's AVs are twice as safe as human drivers, they will allow those cars to operate autonomously as robotaxis. When this happens, the rate of vehicle production is important because the bigger a market share the first mover can capture, the harder it will be for the second company to take any market share, barring significant advantages for consumers to switch to the new player in the market. The amount of time that it takes for the second company to get to the point where their AVs are twice as safe as humans is important, because the longer it takes, the further ahead the first mover can get in terms of safety. If it takes three years for example, the first mover's AVs may at that point already be 10 or 20 times as safe as a human driver, because with an operating fleet of robotaxis their data advantage will increase exponentially. Would you, when you have the choice between a robotaxi from an established company, or a robotaxi from a newcomer that is 5 or 10 times as likely to have an accident, ever go with the second option?

I admit this is some serious hypothesizing, and it's unlikely to play out exactly like this, but the points I'm trying to make are:
  1. Tesla is in a very good position to be the first company to create scalable robotaxis.
  2. There are significant first mover advantages in the future robotaxi market.

Timeline

So far I've only talked about strategies of different companies, and timelines that companies have set for themselves. In this section it's time to look at how close companies are to achieving autonomy, and how long it might take them to get there.


*These statistics are just for California. Some companies also test in other locations.

Most companies report the performance of their AV test programs in terms of disengagement statistics. Considering the average miles per accident in the USA is 165,000 miles, and assuming not every disengagement would've let to an accident, Waymo is not doing terrible. However, none of these companies are working on solutions that are easily scalable worldwide*, being only as safe as a human is unlikely to be enough to convince regulators, and most worrisome is the fact that these companies are going to need many more miles than this to convince them, even when their cars are many times safer than humans. A couple million miles are not going to be enough, they're going to need on the order of billions of miles. Therefore, I don't see any of these companies launching a true worldwide robotaxi service anytime soon. At best they'll launch local ones with safety drivers behind the wheel to show progress to investors.

*And their disengagement data might also be cherry picked. It's no coincidence that Phoenix, Arizona is a popular location for AV testing. The average precipitation days per month is about three. Weather does not interfere with AV sensor suites there.

As I mentioned, Tesla's solution doesn't need to be scaled, because it already works everywhere. And Tesla has ample data to convince regulators when its software has become sufficiently safe. But can Tesla's software ever become safer than humans, and if so, how long will this take?

At the Tesla Autonomy Day Elon claimed that they will be 'feature complete' by the end of 2019, be as safe as humans around mid-2020, and expect to receive regulatory approval in at least some jurisdictions by the end of 2020. These are some wild claims, and I don't blame you if you don't believe them. Elon's timeframes have been very optimistic in the past, and I don't think things will play out exactly like this either. Besides, Tesla raised money shortly after their investor day, so I suspect Elon may have erred on the side of optimism more than usual to generate some hype.

I believe that by 'feature complete' Elon means that they will add enough software so that Autopilot can be used on every type of road, but will still need to be supervised. I actually think that him claiming that they will achieve this by the end of 2019 isn't all that far off, because Autopilot can already navigate all highways, and Tesla is very close to launching Enhanced Summon which will navigate all parking situations. They've also already demonstrated in-development versions of software that can navigate regular city streets. Even if they miss their self-imposed 2019 deadline, I'd be surprised if Tesla Autopilot wasn't able to navigate all roads by Q1'20 or Q2'20 at the latest.

Where I do diverge quite a bit from Elon, is his notion that thanks to an 'exponential rate of improvement' they will go from feature complete to being safer than a human in just six months. Elon understands the road ahead better than I do, and in recent years he has been less overly optimistic than he used to be, so a surprise isn't completely out of the realm of possibility, but I'd expect it to take anywhere from a year (still optimistic) to a few years. It's hard to know exactly, because Tesla only reports safety numbers and not disengagement statistics, so all we have to go by is anecdotal evidence from customers who use NoA. There are tons of impressive stories of Autopilot preventing near accidents, but there are also plenty stories of customers needing to take over because of Autopilot mistakes.

All in all it's kind of a guessing game. I think Tesla could solve autonomy in under two years, but it could take up to five. Either way, I do know that Tesla is on the right track, and when they do solve autonomy, it will be a BIG deal. Uber's self-driving division is valued at 7B$, Mobileye was valued at 17B$ two years ago, GM's Cruise is valued at 19B$, and Waymo is valued at over 100B$. These companies don't do anything but develop self-driving tech, yet Tesla appears to be the clear leader in the industry, has a booming car business to boot, and is currently valued at under 40B$. Something here doesn't seem right to me.

Elon Musk


I'm not going to go in detail on who Elon Musk is in this post. If you want to know more about him, I highly recommend reading his biography by Ashley Vance, or this series of posts about Elon and his companies by Tim Urban on WaitButWhy. Back in 2015 Elon was the main reason I invested in Tesla, and he still is a large part of why I am so bullish on Tesla. But I know there are people who are the complete opposite of me, and who see Elon as a reason to be cautious about or even bearish on Tesla. So I feel like I should explain why I see Elon as such a huge asset to investors.

The Good

Elon does not quit. This is a man who in 2008 in the middle of a divorce, while under constant ridicule and attack by the media, when it looked like both of his companies would fail, and it looked like he would lose his entire fortune in the process, still did not break. Literally his entire life was falling apart, yet he did not give up. Instead, he worked almost every waking hour ignoring the immense pain and stress, and he saved both of his companies. I know that when Tesla is facing difficulties, Elon is not having dinner with his fellow executives thinking about what he's going to spend his next bonus on, he's doing everything he possibly can to make Tesla succeed.

Talking about success, Elon's list of achievements might be longer than this blog post. He has started and sold not one, but two internet companies (Zip2 and PayPal). He started a rocket company that in just 15 years has outcompeted everybody in the industry, partly thanks to landing rockets which almost nobody thought was possible just five years ago. Elon has made some crazy promises, and more often than not has overly optimistic timelines, but the promises that he has not delivered on are scarce.

Looking at Tesla specifically, Elon (with the help of all Tesla employees) has accomplished some amazing things over the past five years. In 2013 Tesla produced 22,477 cars, in 2018 this number was 245,240, a more than 10-fold increase. Revenue growth has been similarly impressive. Tesla's 2013 revenue was about 2B$, 2018 was over 21.5B$. This CAGR of 61% would be impressive in any industry, but even more so in an industry as capital intensive as car manufacturing. Past results are not a guarantee for the future, but I do think it bodes well.

Last but not least, I want to talk about Elon's integrity. If you want to be rich and enjoy the spoils of war, you don't make 200M$ during the dot com bubble and invest every last cent of it into a rocket company and an electric car startup. Elon's main motivation is not to be rich. This man clearly just wants to make the world a better place, and he is working his ass off to try and do so. Furthermore, if you follow Elon for a while and listen to his interviews, you're able to tell that he cares deeply about doing the right thing for his customers, investors, and humanity as a whole. I find companies like Facebook, banks, and other automotive companies hard to trust, because they seem to value their bottom line more than anything. Just look at Volkswagen and Dieselgate. Elon and Tesla on the other hand, have built up a degree of trust with me that I don't have with any other company or business person. They're not perfect (SolarCity acquisition), and I don't blindly trust them (or anybody for that matter), but for the most part I feel confident in trusting Elon and Tesla to do the right thing for me as an investor, the customers, and most importantly humanity as a whole.

The Bad?

At this point you might be wondering if I slept throughout all of 2018 and missed all of last year's controversy. I didn't, and here's my opinion on the biggest ones:

Q1'18 conference call lash out against two analysts.
During the call Elon was very rude against two analysts, refused to answer their questions, and then spent a good 10 minutes answering questions from a single YouTuber. In spite of being a fan of the YouTuber, I thought it was unprofessional as well as cringey at the time. The questions the analysts asked definitely weren't great, but I still don't like the way Elon handled it. I think Elon also realised that he could've handled the situation better, because three months later during the Q2'18 conference call he apologized to both analysts. A mistake for sure, but we all make those and this is not something that influences my opinion on Elon, nor my outlook on Tesla.

The pedo tweet
Last year during the Tham Luang cave rescue in Thailand, somebody tweeted at Elon asking if he couldn't help the rescue effort in any way:



Eventually Elon was put in touch with the leader of the dive team, who told him his help would be very useful*:

Email Exchange between dive team leader Richard Stanton and Elon:

*Admittedly there are numerous people involved with the rescue operation who said after the fact that the submarine Elon built was too big, but I don't see how Elon could've known with such few instructions, and on such a short deadline.

So Elon, one of the busiest people in the world, who just spent the last half year working up to 20 hours a day, and sleeping on the floor in a conference room in the factory to ramp Model 3 production, sacrificed what was probably his only bit of free time that week to design and build a submarine, uses his personal funds to send it to Thailand, and what does he get in return?




He receives public ridicule, and is basically called an attention whore. Then this British dude goes on TV and during an interview further ridicules Elon, and tells him he can "stick his submarine up his ass". Elon in response calls him a "pedo guy" on Twitter. Realising his mistake, Elon deletes the tweet shortly thereafter, but it was already too late. Everybody was mad at Elon.

When I found out, I was ******* livid. But not at Elon, I was livid at the injustice of what had transpired. I admit Elon should not have called him a pedo guy, but I'm hard-pressed to think of anybody who's done more for humanity than Elon. The appropriate response to this very understandable mistake should've been to point it out, and then to forgive it. I'm way more disappointed in society's response to what Elon did, than in Elon for saying something in anger that he shouldn't have. The 'pedo tweet' changes nothing about my views when it comes to Elon or Tesla.

If you're interested in reading more about these events, there's an amazing Quora answer about Elon and the Thai cave rescue.

Funding secured
In August of 2018 Elon was considering taking Tesla private, and talking to various investors about making it happen. In an attempt to be transparent, Elon sent out a tweet that has since become infamous.



I personally interpreted this tweet exactly as Elon meant it. It clearly states that he is considering taking Tesla private, and that he has enough funding to do so if he so chooses. However, others including the SEC, didn't interpret it the same. The SEC sued Elon and Tesla, and claimed that he had tried to manipulate the stock price with false claims.

I've heard a lot of people argue both ways, including an ex-SEC employee who thought Elon did nothing wrong, but the majority of people seem to think Elon was in the wrong. Like I said, I personally interpreted his tweet correctly, and appreciate him trying to be transparent. I think he could've worded it slightly better along the lines of "Funding not a problem.", but that's it.

A settlement was eventually reached forcing Elon and Tesla to pay a small fine, Elon to step down as Chairman of the board, and putting in place restrictions on what Elon can and cannot tweet. Overall not a big deal, and for me nothing changed with regards to my views about Elon and Tesla.

I <3 Elon

I think we as humanity in general, and Tesla investors specifically, are lucky to have Elon. He's incredibly smart and talented, and is dedicating his life to create a brighter future for humanity. Elon was and still is a big reason of why I'm very bullish on Tesla. Tesla's toughest times may be behind it, and a certain degree of success has been achieved, but it'll have to accomplish much more to deserve its current valuation. And to live up to its full potential of being the biggest company on earth, it needs people like Elon to continue to execute extraordinarily well.

Risks


As with any investment, there are risks associated with investing in Tesla. However, part of the reason why I am currently so bullish on Tesla, is because I think the risks at Tesla's current valuation are very small. But either way, here are my biggest concerns when it comes to my investment in Tesla.

Elon

I don't think of Elon as an asset and a risk, I think of Elon as a risk because he is an asset. If something were to happen to Elon, Tesla and all the reasons to invest in Tesla that I talked about would still be here. However, Elon is an integral part of Tesla, and without him I'd be a lot less certain in their ability to execute during difficult periods.

Elon only just turned 48, but he works an insane amount of hours (rumoured up to 90-100 per week during busy times), and he deals with immense amounts of stress. Fortunately enough modern medicine is quite good, and the chance that something bad happens to him in the next decade health wise is not that high. Regardless, one of the most positive pieces of news I could imagine coming out about Tesla, would be that they found a suitable COO ala Gwynn Shotwell at SpaceX to take some responsibilities away from Elon.


Earthquake

Call me crazy, but California where the vast majority of Tesla's operations are located is due for a huge earthquake. It's probably not going to happen anytime soon, but it might, and I'd like to hear somebody ask Tesla during a conference call what precautions they've taken to make sure the Fremont factory is earthquake resilient.

Tesla Service

This is more of a short term worry than a long term risk, but Tesla has been losing significant amounts of money on Service as of late. In 2017 they lost 228M$, and in 2018 they lost 489M$ on Service alone. If nothing is done, I could see this growing to nearly a 1B$ loss in 2019. Tesla has never set out to make a profit from Service, but their goal isn't to lose money on it either. So what is up with this, Elon? I'd love for this to discussed during an upcoming conference call.

Another problem with Tesla's service is that it is the worst part of their customer experience. On online fora you'll find that even the most loyal supporters are sometimes frustrated with certain aspects of their interactions with Tesla. The servicing of cars, and the delivery process seem to have improved somewhat as of late, but used to be pain points. Most importantly, from what I can gather communicating with Tesla seems like a major headache. There are many stories of slow response times, no responses at all, and bad communication in general from the side of Tesla. As a company grows larger and larger it is inevitable that some percentage of customer experience is going to be bad, but it definitely seems like Tesla needs to spend some resources on improving this part of their business.

Margin of error

Currently Tesla happens to be sitting on a huge pile of cash, but at times they've come surprisingly close to bankruptcy. Excluding 2008, Tesla came close to bankruptcy in 2013, and close-ish again in 2018.

In early 2013 just after the launch of the Model S, not enough customers were converting their reservations into orders. Tesla had just ramped up production, and was spending a lot of cash as a result. Tesla got so close to bankruptcy that Elon talked to his friends, Google Co-Founders Larry Page and Sergey Brin, and had a verbal and handshake deal for Google to acquire Tesla. Just before the deal was supposed to take place however, Model S orders started flowing in, the deal was no longer necessary, Tesla had a record quarter, and six months later the stock had gone up from 30$ to 180$.

In an interview in late 2018*, Elon admitted that Tesla had been "within single digit weeks of bankruptcy" in early 2018. In essence, Tesla had made some mistakes in designing the Model 3 production line, and an important supplier had majorly screwed up, resulting in a much slower production ramp than expected. This caused Tesla to bleed cash and come closeish to bankruptcy, but fortunately enough Elon and his team were able to solve the issues in time.

*Here is the 1 minute part where he talks about the bankruptcy risks at the start of 2018.

I think it's highly unlikely we'll see anything like these situations happen again though. Partly because Tesla has learned from their past mistakes, and is continually improving over time. But more importantly because Tesla's established business has gotten stronger and stronger.

When Tesla went from producing less than 1,000 Roadsters per year in 2011 to producing more than 20,000 Model S sedans per year in 2013, this was very risky because they could not fund this expansion themselves, and if the Model S hadn't sold it would've bankrupted the company in no-time. Going from 100,000 Model S and X in a year in 2017 to 100k S+X and 150,000 Model 3 in a year in 2018, still required them to raise capital, and would've bankrupted Tesla if they had been unable to ramp up Model 3 production, but the margin of error was much larger.

Just as it's not that risky for Toyota to launch a new vehicle program to expand production from 10M vehicles per year to 10.5M vehicles per year, launching new vehicles and expanding production for Tesla is becoming increasingly less risky and easier to finance. The Gigafactory 3 in Shanghai for example they've been able to fund entirely through local debt with favorable terms. On top of this Tesla's expansions for the next few years are the further ramping of Model 3 production, and the launch of the Model Y which is an extremely similar car to the Model 3, so I don't see a whole lot of risks there.

Recession

Although I would consider myself an expert when it comes to Tesla, I am not very knowledgeable about the larger economic climate. I have heard people, including Elon, talk about the possibility of a recession in the not too distant future, which brings problems for all companies and Tesla is no exception. If a recession were to happen, Tesla's stock price would obviously drop, potentially a lot because it's a very volatile stock. It would also undoubtedly impact demand, but because Tesla is currently supply constrained, I don't think this would have a material impact on their business.

A recession would also make it harder for Tesla to raise funds, but they are currently sitting on a very large pile of cash thanks to their recent capital raise. Furthermore, with the successful ramp of the Model 3, they are now bringing in a lot of cash, so overall I am not too worried about bankruptcy during a recession. I think the worst that would happen is that Tesla might have to slow down its growth.

Profits

Tesla is frequently criticized for their lack of profits, but I don't see a problem here. Tesla is a high growth company (revenue CAGR of 61%) in a very capital intensive business, and is planning to be many times bigger than it currently is. Tesla is setting up the company and its infrastructure as such, and is heavily investing in growth. As a result their profits have been lacking, but Amazon used to be the exact same and look at where they are now. As Amazon's business has matured, profits have followed. I believe the exact same thing will happen to Tesla, and as a matter of fact, it's actually already starting to happen.

As of the second half of 2018 Tesla has started showing semi-consistent profits. In the second half of 2018 Tesla's EBIT + R&D (Earnings Before Interest and Taxes minus Research and Development costs) was 1.5B$, for a yearly run rate of 3B$*. This shows that although they're still in super growth mode, they're already starting to show decent profits from their core business operations.

*I don't think this is quite sustainable, but I do think 1.5B$ or 2B$ in EBIT + R&D is very sustainable at the moment. But before we'll be able to find out for sure, I think Tesla will have expanded and grown bigger.

Enemies

If you follow Tesla in the media, you will hear about a lot more things that are supposedly wrong with Tesla. Besides a lack of profits, you will often hear that Tesla is going bankrupt, there is no demand for their cars, that Elon is a pathological liar, that the competition's 'Tesla-killers' are just around the corner, that the cars are not safe and have a mountain of issues, etc. etc. etc. I have already covered why all of these are incorrect, so then why is the image that the media creates of Tesla so wrong? Everybody has critics (which is a good thing), but in the case of Tesla something seems out whack.

Negative Tesla headline from October 2009*:

*I actually had to double check the date on this article, because I couldn't believe my eyes. The e-tron was apparently first shown to the public in 2009, but it took Audi 10 years to put it into production. Audi only just started shipping this car, and they're having enormous difficulties producing more than a few hundred cars per month.

I left this one picture up, because I think it's super funny, but my plan was to actually put up about ten pictures of unfair negative Tesla headlines throughout the years. What I found out however is that there wasn't actually all that much negative news on Tesla until about 2016. For the year 2011 I actually couldn't find a single article that I thought was unreasonably negative about Tesla. I even tried to search for "Tesla bankrupt" specifically, but between 2009 and 2015 I couldn't really find any news suggesting Tesla would go bankrupt. The closest I found were stories from 2015 reporting that Tesla had nearly gone bankrupt in 2013, because that's the year when Elon's biography came out that talked about the near bankruptcy and near buyout by Google in 2013.

From 2016 and on wards however, the story is completely different. If you search Google News for stories about Tesla and bankruptcy after 2016, you will find numerous articles claiming Tesla is headed for bankruptcy. The weirdest thing about all of this is that the risk of Tesla going bankrupt has decreased over time, not increased. So what is going on exactly?

A possibility is propaganda coming from some of Tesla's enemies, most likely people involved with the oil industry, perhaps also the car industry. Although I have a feeling that the automotive industry is more likely to have underestimated Elon and Tesla, rather than to have consciously launched a propaganda campaign against them. The same cannot be said for people from the oil industry however, because The Koch Brothers (billionaires involved in the oil industry) launched an assault on electric vehicles in 2016, reportedly with a war chest of up to 10M$ per year. Tesla is the company that has single-handedly pushed the industry towards EVs, and the timing coincides with the steep increase in negative articles on Tesla.

Another plausible cause was first brought to light by Tesla Motors Club (TMC) member "jesselivenomore" in July of 2018. In his lengthy forum post, that I highly recommend any Tesla investor to read, he explains how a group of short sellers (including Jim Chanos) has been strategically waging war on certain companies dating back to 2002. They attack companies that are reliant on financing, such as financial services companies, and try to bring them down by cutting off their access to said financing through a series of media assaults.

Jesselivenomore goes on to explain that this group of short sellers started betting against SolarCity in 2015, because they believed that the debt SolarCity was accumulating through the "no money-down solar leasing" business model, and SolarCity's reliance on raising new debt to continue their growth, made SolarCity a prime target for their attacks. In 2016 when Tesla acquired SolarCity, Chanos and his short seller friends moved their short positions onto Tesla, the exact same time articles surrounding Tesla and bankruptcy started appearing amass. When you google for "tesla chanos" in 2014 there are no relevant results, when you google in 2015 there are less than a handful, and when you google in 2016 there are too many to count.

To me it seems clear that there are entities (like Jim Chanos and the Koch Brothers) who benefit when Tesla's stock goes down, and try to manipulate it in their own favor through the media. Recently it's gotten quite out of hand, and there are even Tesla supporters claiming that Wall Street analysts are in on it too. Whether there is any truth to those accusations or not I don't know, and I think bulls need to be careful of falling into a pattern of brushing off all negative news as conspiracies against Tesla. The timing of certain analyst's reports do seem suspicious at times, but they could be coincidences and there's no proof of manipulation by analysts.

The reason I am talking about all of this in the Risks section of this blog is that I believe that the powerful entities threatened by Tesla, and the negative narratives that they are pushing in the media pose a certain threat to Tesla. Like I mentioned when talking about Tesla's brand, I think Tesla's brand image has taken a hit over the last few years, partly due to some mistakes Elon made last year, but moreso because those mistakes have been unfairly accentuated by the press, and because of all the increasingly negative narratives surrounding Tesla.

Fortunately enough Tesla also has a growing base of very loyal customers and supporters, but it can't be denied that the negative narratives are hurting Tesla's image. Whether it is enough to really damage Tesla's reputation and business, I don't know. But the fact that I can't rule it out, is why I see it as a risk to my Tesla investment.

Overview of the risks

I don't think that an earthquake is likely to seriously threaten Tesla, and think that in a worst case scenario it is more likely to just be a disruption. Tesla's service losses are a short term worry, not a long term risk, and I am confident Tesla will eventually get around to improving its service and customer experience. It has been brought up during a recent conference call, and Elon said it was his utmost priority at the time. Hopefully things will improve in the near future.

There's nothing I can do about the risk of something happening to Elon. All we as investors (and humanity) can do is hope he turns out to be an alien after all, preferably an immortal alien. Besides, even if something happens to him, it's not certain that it would significantly impact Tesla.

I mentioned that I don't buy into the theory that Tesla cannot turn a profit. With continued growth (and hopefully soon autonomy) I think it's just a matter of time before they will generate large profits. I don't think a recession poses an existential threat to Tesla either, most likely it would just slow them down. In terms of there being a small margin of error, I explained that the worst should already be behind us. There's always going to remain some risk in any business, especially a high growth one like Tesla, but I have tremendous faith in Elon and his team, and their abbility to continue to execute like they have in the past.

Tesla's powerful enemies and their media attacks might be the biggest concern I have about Tesla's future. Nonetheless I remain optimistic because of everything I've covered so far. I believe that Tesla is a company with extremely solid fundamentals that is poised for success.

Financials


I think by now I've made it clear that Tesla has tremendous future potential, but nobody would buy the stock if Tesla was valued at 500 trillion $. And if Tesla stock was free, everybody would love to get their hands on some. What I'm trying to say is that, I can't argue that Tesla is a good investment without talking about its financials and valuation, so here goes.

*Google Drive links to the models I discuss in this section can be found at the very bottom of this blog post. If you're on a PC, it might be easier to follow along by opening the actual spreadsheets, rather than looking at the images.

Where Tesla is at

Tesla's financials 2013 - 2018:
*Current share price as of the end of June 2019 is 220$ for a Market Cap of 39B$
**All $ values except for share price are in thousands

Back in 2013 when Tesla had just launched the Model S, it was a tiny company that produced just 22,000 cars in a year and brought in a total revenue of 2B$. Tesla's revenue multiple was 8.9, and even Elon Musk himself made comments that he thought Tesla was valued higher by the public markets than they deserved. But since then Tesla's growth as a company has significantly outpaced the growth of Tesla's stock price and market cap.

If I gave you the option to invest in one of the following four companies, which one would you choose?

  1. 2013 to 2018 total revenue growth of <10%, and a revenue multiple of ~0.75.
  2. 2013 to 2018 total revenue decline of <10% and a revenue multiple of ~0.35.
  3. 2013 to 2018 total revenue growth of 20-25% and a revenue multiple of ~0.30.
  4. 2013 to 2018 total revenue growth of 800% and a revenue multiple of ~1.65,

Your best answer would be to tell me you want more information about all four companies, but as is company #4 looks pretty damn cheap, right? Number 4 is of course Tesla, the other three I randomly selected and are: #1 Toyota, #2 GM, and #3 Volkswagen.

I'd love to be able to scientifically prove to you the value of Tesla as a company, but sadly company valuations are not an exact science and ultimately depend on what the overall market values it as. The best I can do is make comparisons with other companies, and show you future projections. Regardless, I think it's hard to make an argument for Tesla being anything but undervalued at the moment, unless you buy into some of the "bankruptcy is imminent" and "there is no demand" narratives being pushed by the media. The facts are that Tesla is a hyper growth company, that is barely valued more than other companies in the industry that are more or less stagnant. On top of that no one seems to put any value into Tesla's autonomy business, in spite of Tesla being the leader in this space. Two of its competitors have been valued at ~20B$ (Cruise and Mobileye), and one of them at over 100B$ (Waymo), whereas Tesla's total valuation is ~39B$ including an automotive business that could be considered more than 39B$ on its own.

The only argument against this that has some validity is that Tesla has never had consistent profits, which is ultimately what is most important. Tesla's current EBIT + R&D multiple is approximately 25*, compared to Toyota and GM's which is approximately 7-10**. Considering what I said earlier about Tesla setting itself up to be a much bigger company than it is today, I don't even think this is bad. Looking at their EBIT + R&D as a percentage of revenue, Tesla's is currently at about 7%. Amazon used to be in a similar situation around 8-9% in 2011 and 2012, but has since increased this to ~15% in 2016 and 2017***, and some might argue Amazon is still a high growth company that still hasn't reached maturity yet.

*Assuming yearly EBIT + R&D of only 1.5B$. 2018 was 1.2B$, but second half of 2018 was 3B$.
**Hard to find out precisely, because apparently they don't publish their R&D expenditures.
***2018 was a dip year at 12%

I believe that if you look at Tesla in a complete vacuum and consider neither paste nor future, they are overvalued. But if you take a look at their past and take into consideration that they are growing very fast, it's hard to argue against Tesla being undervalued and there being a chance of modest returns. When you start to look at Tesla's future, it becomes harder to predict what will happen, but the degree to which Tesla can succeed also starts to become larger and larger.

Where Tesla's EV business is going

My Tesla Bear Model for 2017-2030 (Automotive):

*Revenue and Profit numbers are in thousands
**The highlighted sections are where I made assumptions

I think these are very conservative numbers with Year over Year (YoY) growth percentages between 16% and 48% until 2025, compared to the 35% - 138% annual and 61% compounded annual between 2013 and 2018. These numbers also assume a Model Y ramp that is barely any faster than the Model 3 ramp, in spite of the cars being extremely similar and Tesla being able to build upon lessons learned from the Model 3. Furthermore, it assumes the Semi Truck going into production in 2021 although Tesla is still aiming for 2020 (I'm skeptical), and the Pickup Truck not going into production until 2023, even though I could see them announcing later this year that production will start as soon as 2021. This model is even more conservative after 2025, where it assumes steadily declining growth rates of 18% to 10% per year.

My Tesla Bear Model for 2017-2030 (Financials):


So if you consider very conservative automotive growth, take all other parts of Tesla's business out of consideration completely, account for dilution, and value it as a normal car company in 2030, Tesla's stock would still be worth more than 4 times what it is today.

Now let's take a look at less conservative numbers.

My Tesla Bull Model for 2017-2030 (Automotive):




Here I am assuming much more aggressive growth until 2025. These numbers still assume less growth than Tesla achieved over the last 5-6 years, but as a company grows larger and larger it becomes harder to grow at the same pace. I could see them missing some of these for sure, but I also think it's possible they will hit these numbers or perhaps even slightly exceed them.

This model too assumes very small growth after 2025. Considering Tesla has such a lead in EV tech, I think the 2030 number for a production of 5M is still rather conservative, but the longer into the future I predict, the more chance there is that I am wrong, so it's good to err on the side of caution a bit. Especially if Tesla manages to turn their factories into a product, and learns how to expand their production at scale, I could even see them producing 10-20M cars per year by 2030 in a very bullish scenario. I think this is very unlikely, but not out of the realm of possibility.

My Tesla Bull Model for 2017-2030 (Financials):


Taking these more optimistic (and in my opinion more realistic) predictions, we can start to see the potential of Tesla's automotive business. Again I've taken out all non-automotive parts of Tesla's business and accounted for dilution, but this time I'm valuing it differently after 2025. I'm valuing it on an EBIT multiple of 30, which is the exact same way the market currently values Alphabet*. In this scenario Tesla would be worth about 15x in 2030 compared to what it is today.

*I'm aware you can't value a company by randomly picking another company, taking the way the market values it, and applying it to the original company. But there is no scientific way to do this. For comparison's sake, Toyota's EBIT multiple is a little over 10, and Amazon's is at ~80.


Where Tesla's Autonomy business is going

Okay, so this is where things are going to get nuts, and I fully expect that some people will think I'm crazy after finishing this section, because of the absurdity of some of the numbers. I'm going to continue with the same two models I used above, but add autonomy to them.

My Tesla Bear Model for 2017-2030 (Autonomy):

*Revenue & Profit in thousands

Let's first go over how I got to some of these assumptions:
  • I'm assuming here that Tesla will start to get regulatory approval in 2024, four years after Elon's estimation. It could definitely happen sooner, it could definitely happen later, but to be honest it doesn't impact the model much. FSD can be turned on with a single Over-the-air (OTA) update after all, so what matters is the number of FSD capable cars on the road.
  • "FSD Cars" are cumulative, and taken from the automotive model in the previous section.
  • "% on network" is the percentage of cars that are on the network. I think 80% long term is very conservative. Would you let your car rot in the garage, or let it make thousands of dollars a month for you with the press of a button?
  • "Avg hrs / week" is the time cars work on Tesla's network each week. Elon estimated 16 hours a day in his presentation, this model is much more conservative at 12 hours per day.
  • "Avg MPH" is also very conservative. Elon used 16MPH in his presentation, but the average Uber trip speed in the US for Jan-March 2015 was 24.9MPH, for Oct-Dec 2015 it was 27.0MPH.
  • "Useful miles" is the percentage of miles with a paying customer. 50% is the same as Elon's assumption, but less than what some Uber and Lyft drivers report. I found one article claiming about 1:1 ratio of useful vs non-useful miles, and another claiming 69 non-useful miles for every 100 useful miles.
  • "Cost / mile" Elon claimed is at <0.18$ / mile today, so my guess seems conservative to me.
  • "$ / mile" is the price per mile. Uber and Lyft charge 2-3$ / mile, Elon used 1$ / mile in his presentation, but admitted he just took it out of thin air. Tesla could potentially charge more initially, because Uber and Lyft are so much more expensive, but if eventually there are robotaxi competitors they might have to drop prices. Nobody can really predict this, but this is what I went with.
  • "Tesla Cut" is Tesla's cut of the profits. I believe current ride-hailing apps take 25-30% of total revenue, and Elon also suggested they'd take 30% of total revenue in his presentation. The 50% of profits I am assuming works out to 20-25% of total revenue, so it's quite conservative in my opinion. I have Tesla's Cut increasing over time, because Tesla will also operate their own fleet of cars from which they receive 100% of profits. The 65% works out to Tesla having 30% of 80% of 21M = ~5M robotaxis in their own fleet in 2030. This might be a bit optimistic actually, but even with a cut of 60% or 55%, the numbers wouldn't change dramatically.
My Tesla Bear Model for 2017-2030 (Financials including Autonomy):

I've again taken out all of Tesla's Energy business for simplicity, and I'm assuming an EBIT multiple of 20, which is in between Toyota's 10 and Alphabet's 30. Even though I think this would be an extremely low EBIT multiple in this scenario (remember Amazon is at 80), Tesla's market cap would be nearly 2T$, and the stock price would increase 35-40x. And remember, these are for the most part very conservative assumptions. Before I comment any further, let's go take a look at less conservative projections, which are going to look simply stupid.

My Tesla Bull Model for 2017-2030 (Autonomy):

I've adjusted some of the assumptions here slightly. 100 Hrs / week is still less than Elon's assumption, 20MPH is still <80% of the average Uber ride, and cost / mile is (at least initially) still higher than what Elon says is the cost to operate a robotaxi today.

The craziest number on this entire page in my opinion is neither the 1T$ in yearly revenue in 2030, nor the 425B$ in gross profits, it's actually the fact that only ~1.5 trillion useful miles per year (3T total x 50% useful) are required to get to these numbers. From what I've been able to gather, all 2 billion cars on the road today collectively travel about 20 trillion miles in a year. This means that Tesla would only have to gather about 10% of the 2030 robotaxi market* to attain these already mind-boggling numbers.

*Not 100% travel over roads might be done in robotaxis, but then again global miles traveled on roads will also likely increase between now and 2030.

My Tesla Bull Model for 2017-2030 (Financials including Autonomy):

And yes, that is a 20,000,000,000,000 aka 20 trillion $ market cap off of a 10% market share in robotaxis, raising stock price by almost 400x. Go ahead and call me crazy, I think this stuff is absurd too.

But in all seriousness, let's take a look. I've again taken out Energy for simplicity. I'm actually unsure about my SG&A assumption. In all of these models I've assumed SG&A to be 12.5% of automotive revenue similar to what it's been in the past, but I don't see their SG&A expenses growing nearly that much off of Tesla Network, which is just software. I've assumed 5% of Network revenue in SG&A costs, but I'm honestly not sure about this. In terms of valuation, I went with an EBIT multiple again (after 2025), but this time one sort of in between Alphabet's of 30, and Amazon's which is at 80.

I've also done a financial projection of what would happen if Tesla somehow some way ended up with a monopoly on self-driving cars and robotaxis, more as an exercise than as something that I believe will happen. I'm not going to get into that here though, because I fear I'd lose all credibility, but let's just say Tesla's revenue would come to rival the freaking US economy. I think at a minimum all of this goes to show that robotaxis are a gigantic opportunity (iirc transportation is ~10% of global economy after all), and that it's no wonder that some of the companies in the space are seeing very high valuations (Waymo >100B$).

To round all of this off, let's look at a non-fully exhaustive list of things that have to happen for this to become reality*:
  1. Tesla has to be successful at further scaling both battery and vehicle production to at least a couple of million units a year, in order to create a large fleet of robotaxis.
  2. Tesla has to solve autonomy, and get regulatory approval.
  3. Tesla has to successfully launch a ride-hailing service, and create whatever infrastructure they need to support this.
*Listing every little thing would take way too long, but these are the three biggest in my opinion.

And that's really more or less it. Don't get me wrong though, all three of these are very very difficult, and will require top notch execution from Elon, his team, and all Tesla employees. But honestly, I don't see how #1 is more difficult than what Tesla has accomplished over the last 10 years, I think it's easier. As for #2, this is not a certainty and a hard problem to solve, but is it harder than landing rockets? Basically nobody thought landing an orbital rocket was possible 10 years ago, whereas in the case of autonomy there are actually plenty of experts who believe it's just a matter of time. And as for #3, this will also takes a lot of effort and isn't something Tesla will do overnight, but it is something they can do.

Final Thoughts


This post has become a lot longer than I initially expected it to be. I started writing this over 150 hours ago, and have spent almost every waking hour working on it.

It's probably wise to put in a disclaimer that none of this should be considered as investment advice etc. I'm not a financial advisor in any way, I'm just somebody who's very passionate about Tesla, and has been following the company for the past 4 years.

The most important thing is to always think critically for yourself, and make your own decisions. You can use this post as part of the information you base your decision on, but I recommend you to seek out other sources as well.

If after you've critically thought about all this, you come to a different conclusion than I have, please share your views with me. Critical feedback is so valuable that it's probably worth more than TSLA stock (for now), so I'd love to hear where, and most importantly why you think I'm wrong.

If you've read this far, thank you very much for taking such an interest in Tesla, the most exciting thing going on in our world today.
P.S. Happy 48th Birthday Elon!


Bear Model PDF Download Link
Bear Model Numbers Download Link
Bear Model Excel Download Link

Bull Model PDF Download Link
Bull Model Numbers Download Link
Bull Model Excel Download Link

Comments

  1. paulJuly 3, 2019 at 12:58 PM

    "20 trillion $ market cap off of a 10% market share in robotaxis" We are decades away from self driving cars, and in all likelyhood Tesla will go belly up long before they become a reality. Remember: Tesla (and others) is working with an extremely primitive tool called a neural net, which is not capable of abstract thinking or understanding causal relationships. It only sees pixels arranged in a certain form. A human has to label the pixels and identify them as humans, cars, trees etc. After that labelling the car still does not understand what a human, a car, a tree actually is or how they behave/engage with the environment. However it is hoped that through pattern recognition and rule formulation by humans the car will drive itself. The problem is, is that the world is too complex to write rules that will solve all relevant situations you are confronted with in your daily commute. You need real intelligence that asses the situation in real time. A neural net+written rules are not enough.

    All I am trying to tell you is that we are still a long way off from achieving the goal of a self driving car. To assign any value to Teslas software today is absurd, not only because we are still a long way out from solving this issue in general but also because Tesla is a laggard not leader in this area.

    ReplyDelete
    Replies
    1. FrankJuly 4, 2019 at 2:30 AM

      Thanks for sharing your thoughts, Paul.

      Neural nets can do way more than what you're describing, and Tesla has a number of examples of that in their Tesla Autonomy Day. One example of this is when they talk about cut-ins:

      https://www.youtube.com/watch?v=Ucp0TTmvqOE&feature=youtu.be&t=7819

      In this example, Tesla feeds video into a neural network, and the neural net has learned how to detect when a vehicle is going to cut-in.

      During their Autonomy Day they also talk about how the neural net is learning how to do more and more, and how it's eating into their code base. Although they also mention that certain things are more computationally efficient to do heuristically, so it sounds like some written rules will always remain.

      I agree with you that it could still take quite a while before autonomy is ready, and that it's not a certainty that this will happen in the next decade. However, I think more likely than not it will be ready by 2030 (possibly a lot sooner), and therefore we can assign a value to it. Especially in a bull case.

      Delete
    2. paulJuly 4, 2019 at 4:23 AM

      The problem is that pattern recognition via labelled video/images fed to the neural net is not enough to get the car to drive even close to as safe as the best human drivers. The tool is too primitive to make enough distinctions between situations to safely drive in the real world. For example: If you drive past a pedestrian you know instantly if he is aware of your car or if he seems distracted by looking at his face, body language, estimating his age etc etc. The neural net, since it is incapable of abstract thinking or understanding causal relationships, can only work off a very imprecise general rule. It doesnt comprehend enough to accurately asses the situation. Another example: If you have a 4 lane wide roundabout (ex. Paris Arc de Triomphe), it is impossible for a neural net to safely enter that roundabout, since it cant communicate effectively with other driver. It has to have an aggressive but safe driving policy to seamlessly enter the roundabout. The robotaxi will either be too conservative and thus never enter the roundabout or be aggresive but unsafe and crash into the car. A human driver can assess whether there is an opportunity to enter the roundabout by looking at the other driver.

      Delete
    3. NateJuly 4, 2019 at 7:57 AM

      tldr; I agree with Paul that self-driving AI will not have the capability of abstract thinking ("general intelligence") in the near future. But I agree with Frank that neural nets can accomplish a lot more than Paul is suggesting. In the end, I think we are still 5+ years away from level 5 FSD (no person in the car and no remote access needed to take over when the car encounters a problem).

      Paul, I think that vision neural nets, along with some combination of hard-coding and neural nets for driving policy, do have the potential to tackle both of the examples you bring up: 1. Guessing a pedestrian's intention to enter the road and 2. Merging into a busy roundabout. In both cases, it is "simply" a matter of collecting and labeling enough training data. No abstract thought is necessary, just as it was not necessary for google's AI to win at Starcraft despite needing to at some level predict the actions of a human opponent as well as the response of humans to the AI's actions.

      Getting training data for problem 1 is more straightforward, I think, using a very similar method to the "cut-in" method mentioned by Frank. Tesla vision networks can already identify where the road is and where pedestrians are, so Tesla needs to simply ask its cars to identify times when pedestrians enter the road, potentially cutting off the car and send back video data from ~30 s prior. They also need to send back random examples where pedestrians are near the road but do not enter the road. Then they ask a neural net to "predict" from prior video frames whether the pedestrian will enter the road. The neural net will also "know" the car's speed and any other information from the visual scene, such as the presence of a crosswalk or traffic light. Because Tesla always knows the correct answer (well, this depends on the accuracy of their pedestrian identification and position neural net), these data are always pre-labeled, so a large amount of labeled data can be collected in order to train the network to predict pedestrians entering the road. With Tesla's new hardware, the neural net uses full-resolution video. The video is not HD, but it probably contains sufficient information for a human to judge if the pedestrian is distracted, if he/she is looking at the car, what the approximate age is, etc. (all of those things you mention above). If that information is in the image/video and it is relevant to the problem, then a neural net CAN and likely WILL use that information for its prediction, with enough training data. No higher cognitive reasoning or rules-based programming is necessary. This is one of the very powerful things about neural networks.

      This initial training would probably result in a lot of false positives. This is where the sometimes-maligned "shadow mode" is extremely useful. This neural net could be deployed to Tesla vehicles, and predictions of pedestrian "cut-ins" could be recorded without using the predictions to drive the car. The videos associated with these predictions could then be sent back to Tesla along with the final result (did the pedestrian enter the road or not). This allows Tesla to collect lots more examples where a pedestrian might look like he/she is going to enter the road but does not. Incorporating these examples in a new round of training will help reduce false positives. This process can then be repeated until false positives and negatives are low enough to deploy.

      Delete
    4. NateJuly 4, 2019 at 7:58 AM

      For example 2, entering a big roundabout, I think that the solution will be more complicated but is still tractable in principle. This is just my speculation, but I think that initially Tesla will use some kind of imitation learning. A neural net (or perhaps a rules-based algorithm) will be given examples of human Tesla drivers entering roundabouts and try to predict the behavior of a human. This will include whether or not to "edge forward" to show intention to merge as well as the timing to fully commit to entering the roundabout. Again, with enough data (could be raw data or could be higher-level data like the location and speed of cars in the roundabout), a neural net could be trained to predict what a human would do. If a neural net is used, for extra safety, there also probably should be some rules-based check as well. If a car is traveling too fast to stop, then do not enter the roundabout even if the neural net suggests it. There would also need to be some careful human vetting of false positives (where the neural net recommends entering the roundabout but the human driver does not enter) to make sure that the neural net recommendation is not clearly dangerous. Once the net is judged good enough to deploy (i.e. the "enter roundabout feature" is enabled), then it can continue to learn from real-life feedback, which initially will come from a vigilant human cancelling autopilot (hitting the brakes). Eventually, as the system gets better and humans learn to trust it (get less attentive), it can continue to learn from close calls as well as the inevitable real accidents. Probably the close calls will be more valuable because they will be larger in number and lots of data are needed to get meaningful improvements. Again, none of this explicitly requires higher reasoning but could implicitly incorporate an empirical understanding of human driver behavior. While I am confident that this is in principle do-able, I really don't know how long this will take to get to a high level of safety, nor whether the current neural net processing "FSD" hardware is sufficient. It may be good enough for simple roundabouts if the feature is released within 1 year (with the driver taking responsibility) and tested for another year with real drivers providing input by cancelling autopilot when it goes wrong (assume several hundred thousand cars with FSD hardware by then). For more complex roundabouts, I'm more skeptical that Tesla will be ready in 2 years. The time it takes to really "get" complex roundabouts probably depends on whether a single "general" solution can work for complex roundabouts around the world or whether different countries and kinds of roundabouts need to be tackled separately. I think that we won't know until Tesla tries, and Tesla (and Elon) probably don't know either right now. I may be wrong, since they are apparently already testing roundabouts, and they actually have already implemented merging onto highways.

      Delete
    5. NateJuly 4, 2019 at 7:58 AM

      I think that where higher reasoning would come in would be "one-off" events that are unique but understandable by a human. One example might be a telephone pole that starts to fall over into a road. Another might be a sign "bridge out ahead" or someone at the side of the road yelling a warning. I think that there will be a long time period where Tesla (and other self-driving cars) are significantly safer than humans in some situations (where data are plentiful and constant vigilance is a big benefit) but less safe or more likely to annoy everybody in other situations (where higher reasoning or "general intelligence") is valuable.

      Back to the point of the blog post, it's an interesting question what value Tesla's FSD software will have during this potentially long period where Teslas can handle most but not all situations better than a human. A lot will depend on regulations and potential backlash from problems (like a stuck driverless Tesla blocking traffic or even an accident). I tend to think that there will be at least a couple of years where Tesla FSD is "feature complete" and Tesla makes some money on people buying FSD for convenience and "wow" factor, but not from ride sharing. Then there will be a period of Tesla "rentals" that drive to you autonomously but you need a license and you sit in the driver's seat and don't need to pay attention (level 3 autonomy?) but need to be ready to take over if the car identifies a problem. During this stage, for fully autonomous driving (nobody in the car), Tesla would probably need infrastructure where they can remotely access the car to tell it what to do if it encounters a problem it cannot resolve (i.e. gets stuck). During this period, the value will be much less than a full robotaxi, because of the extra infrastructure needed and the extra driver/passenger requirements. Tesla may even operate this service a close to break-even if they think that higher utilization will help them reach true (level 5) FSD sooner, along with that magical revenue and profit stream.

      Delete
    6. paulJuly 4, 2019 at 9:16 AM

      "With Tesla's new hardware, the neural net uses full-resolution video. The video is not HD, but it probably contains sufficient information for a human to judge if the pedestrian is distracted, if he/she is looking at the car, what the approximate age is, etc. (all of those things you mention above). If that information is in the image/video and it is relevant to the problem, then a neural net CAN and likely WILL use that information for its prediction, with enough training data. No higher cognitive reasoning or rules-based programming is necessary. "

      This wont be possible. Labelling by a human only makes sense if the neural net can spot the difference between the one labelled object (ex. car) and the other labelled object (tree). The neural net is bound by "pixel" understanding. I.e. the reason it can distinguish between a car and a tree is that there are clear differences in pixel arrangement. It is impossible for the neural net to understand why one young man was distracted and the other guy was not distracted because it is not translatable into the language of the neural net. You cant express distraction versus no distraction via the tool of pixel arrangement. Yes the computer might be able to predict in some cases if a car wants to cut into your lane but only because we are dealing with a language a neural net can understand: speed and direction.

      The same argument applies for the roundabout. The neural does not understand why a human in one case says go and in the other case says stop because it does not have access to the information a human brain gathers from its surrounding. It is bound by primitive input and understanding: Pixel, speed, colors.

      Delete
    7. FrankJuly 5, 2019 at 3:28 AM

      Paul, in your example of a distracted pedestrian, how do you think you and I know that this pedestrian is distracted? I'd argue that we use our vision to recognize that a pedestrian is distracted, because we see him looking away, or he has a cell phone in his hands. Unless you think that humans know other humans are distracted because of invisible waves that we send out when we're distracted, a neural net in a self-driving car can learn to pick up on the same visual signs as we humans do.

      Of course a neural net isn't going to just know a pedestrian is distracted, and it would have to be specifically trained in order to be able to make this distinction. But if it turns out that this is important in order to create a FSD car, we can teach a neural net to be able to make this distinction. Basically, as long as humans have eyes with which they can see they can drive a car, and therefore an AI with a neural network and cameras can also learn to drive a car. It's a complex software problem, but a solvable one and just a matter of time. I think it's looking likely that it'll be solved in the next 2-10 years (2 years very unlikely, 10 years super likely), but exactly how long it will take is up for debate.

      Delete
    8. FrankJuly 5, 2019 at 3:33 AM

      @Nate. Nice explanation! I'm not sure if Tesla will do the remote support thing, they might, but I saw a Nissan presentation at a conference with exactly this. They showed the scenario of a FSD car getting stuck somewhere, and then it called a support person at Nissan who helped it navigate through the problematic area. Could be that this kind of set up is necessary or helpful in the initial stages of launching a robotaxi service.

      Delete
  2. paulJuly 5, 2019 at 4:19 AM

    "Paul, in your example of a distracted pedestrian, how do you think you and I know that this pedestrian is distracted? I'd argue that we use our vision to recognize that a pedestrian is distracted, because we see him looking away, or he has a cell phone in his hands."

    Our brain conciously and subconciously processes a lot of information when it looks at a pedestrian/another person. For example one grandma might look at you and you can see that she is mentally still very fit just by looking at her face (how mentally fit does she look basically). A neural net cant detect this, since its a very complex process and it cant be determined with simple pixel differences.
    Somebody looking away does not mean much. How old is this person? Is there a reason why he is looking away? Etc etc. All of this is impossible for a simple neural net to grasp.

    ReplyDelete
  3. SteveOctober 17, 2019 at 5:50 AM

    Hello.
    I really like your analysis!
    What surprises me in your longterm forecasts is that you are much more cautious than ARK and Elon, predicting 1,7 respectively 3mio cars per year in 2023.
    That means that your super bullish thesis relies mostly on FSD?

    ReplyDelete
    Replies
    1. FrankOctober 20, 2019 at 2:11 AM

      Hey Steve,

      I'm glad you enjoyed it so much!

      You're right that especially the bear model I present in this blog is extremely conservative. I think it's a near certainty that Tesla will at least come within 10% of those numbers.

      Yes, you could say that. I think that if Tesla executes well, they'll likely be one of the biggest companies in 2030 just off of electric vehicles. Growing to the size of Facebook/Alphabet etc. and seeing a 10-20x increase in stock price.

      However, my super bullish thesis that would see Tesla being worth multiple trillions of $s and seeing a 50x or more increase in stock price, relies entirely on FSD.

      If Tesla manages to solve autonomy, produce a very large number of cars, and launch a successful Tesla Network, I think a 1-2T$ market cap is conservative, and a 5T$+ market cap is possible.

      Delete
  4. UnknownJanuary 1, 2020 at 12:54 PM

    Reading the AI ML NN discussion - did nobody think of letting Tesla Owners help with labeling while supercharging? The same way captcha uses humans to label street scenes on the web? Tesla could give those people credits, free supercharging miles or something else?

    ReplyDelete
    Replies
    1. FrankJanuary 2, 2020 at 4:29 PM

      This might be possible, but although it seems likely that Tesla could always use more data labeling, we don't know for certain that this is a large bottleneck for them. They're currently working on "Project Vacation" which is supposed to automate the entire training process, including labeling it sounds like. Details are scarce, but perhaps they've found a different better way to effectively label a lot of data.

      Delete
  5. TuBuiApril 5, 2020 at 6:46 PM

    I throughly enjoy reading this, thank you for the wonderful write up!
    Like you, I became a huge fan after reading Elon biography, it's such a good read that I read it twice and will read more in the future. Then bought TM3 plus range - the best car I ever driven by huge margin. Then I became an investor. Im new to stock market, and my first share I bought is TSLA at $200~ last year. Today, there are 412 shares in my portfolio as I dig deeper into the company and see its potential.
    I agree 95% of your argument and have seen those arguments recently popping all youtube by various youtuber, I wonder where they got their inspiration from =D
    One of things I want to debate with you is your thesis of 80% of all Tesla will be robotaxis. I think 80% is too optimistic.
    1. Most of ride hailing services are in major cities while smaller cities are relatively absent of them. Smaller cities/rural area are small in populations and are too spread out to have reasonable profits for ride hailing services. Source "https://www.pewresearch.org/fact-tank/2019/01/04/more-americans-are-using-ride-hailing-apps/"
    2. USA has complicated regulatory that Tesla have to face, imagine if Tesla have to go through 60 different country's regulatory. Robotaxis may exist in certain area and may not exist in other areas.
    3. Some Tesla owner are car enthusiast and that they dont view their Tesla as an investment but rather something they can enjoy driving and owning. They dont want their object of desire slowly degrading.
    Sorry for my poor English, and again, thank you for your wonderful blog! I usually dont like to read but your blog have captivated me.

    ReplyDelete
    Replies
    1. FrankApril 6, 2020 at 5:53 PM

      Tubui,

      Thanks you for your comment. I'm delighted to hear you enjoyed reading my blog so much! And congrats on your TSLA investment, which looks like it's already paying off :)

      With regards to the points you make:

      3) You're right, and that's why it'll probably never go to 100%, but I believe it'll get quite close. The biggest reason for that is that AVs will be much more expensive than vehicles are today, and a taxi ride in an AEV will be just as cheap as a bus ticket.
      As a result, most individuals will no longer own a vehicle and instead opt to use a ride-hailing service instead. The consumers buying AVs will mostly be fleet operators and people who intend to make money from the AV.
      If a simple AV can earn ~$200k over a 10 year period as a robotaxi, a lot of people are going to want to buy it for $100k as an investment, but not many people will want to spend $100k on a private vehicle when taxi rides are as cheap as bus tickets.

      2) This will definitely be a hurdle, but I think it'll be relatively easy to overcome. At the rate Tesla is gathering real world data, it doesn't take them long to gather billions of miles of data, and with billions of miles of data, that unequivocally show an autonomous Tesla to be safer than a human, it shouldn't be too hard to convince regulators to allow autonomous Teslas to operate in their jurisdiction. It'll save lives after all.

      1) It's definitely easier to operate a ride-hailing service in a city, there's no question about that. There's also no question about the fact that some areas are too remote for ride-hailing, because it doesn't make much sense for someone, who lives 50KM away from his nearest neighbour, to share a vehicle with anyone. He needs his own private vehicle.
      However, even in the vast majority of less-populated areas an autonomous ride-hailing service can work. Perhaps it'll be a little more expensive if the AV has to drive further to pick up new customers, and perhaps waiting times will be slightly longer, but it should work in the vast majority of inhabited areas.

      With all that being said, I could certainly be wrong. Nobody knows for sure how the autonomous future will unfold, but it will certainly be exciting :)

      If you liked this blog, you should also check out the follow up to it:

      https://teslainvestor.blogspot.com/2019/12/my-tesla-investment-thesis-20-teslas.html

      Delete
  6. David ZhaoApril 20, 2020 at 11:28 PM

    Excellent post Frank again!
    I am moving back to Singapore from San Francisco this August.
    Let me know if you need any help in your plan to move from KL to Singapore? I am more than happy to help!

    ReplyDelete
    Replies
    1. FrankApril 21, 2020 at 6:08 PM

      David Zhao, right?

      Thanks! I'm not quite moving to SG yet, but hopefully soon yes :) I'll let you know if I need any help, thanks for the offer!

      Delete
  7. RyanMay 25, 2020 at 10:25 PM

    I will forsure be tuning into future blogs after reading this. I was very enlightened by the Robotaxi math.

    ReplyDelete
    Replies
    1. FrankMay 26, 2020 at 12:57 AM

      That's great to hear, Ryan!

      My big posts take dozens of hours to write, and work/life gets in the way sometimes, so I don't have a regular posting schedule. Often weeks or even a few months can pass in between posts.

      However, I try to do an earnings forecast every quarter before Tesla releases its numbers. Also, there are currently 18 published posts, which altogether should take at least about 10-15 hours to read. So if you haven't read those yet, you could check those out in the meantime.

      Delete
  8. JungyoonJanuary 7, 2021 at 1:40 AM

    Thank you so much Frank for your thoughtful posts. I learn so much from you.

    ReplyDelete
    Replies
    1. FrankJanuary 7, 2021 at 2:44 AM

      I'm glad to hear that, Jungyoon! You're welcome

      Delete

Post a Comment